ABSTRACT
Purpose: Purpose of this research is to carry out a machine learning intelligence based innovative method to determine quality of food which may be hazards to health if consumed by humans. This article detects human sickness by sensing nutrition that causes smells.
Theoretical framework: In developing nations, people just focus on basic need of food rather than focusing on the quality and the nutritional values of food which are exhibiting hazards impact of unhealthy food on the lives of people. Many people are suffering from diabetics, cancers, cardiac problem, liver problems and stomach related health issues which are originated due to consumption of bad food. Consumers are satisfied with food quality, and more individuals are assessing it.
Method/design/approach: As a methodology, an electronic nose uses chemical sensors to identify complicated odors. Standard technologies can detect gases from households, industries, and explosive materials. It cannot fulfill freshness requirements. Electronic noses, computer vision, and other sensory approaches may imitate human olfactory, taste, visual, and sensory qualities, both pleasantly and unpleasantly. Neural networks organize innovative artificial/mechanical intelligence systems to interpret fragrance recordings for human brain recognition. Inspired by human brain processing, we offer optimized feedback, centroid clustering, and self-organizing maps for machine learning systems to identify smell data. This work proposes a simulation technique based on benchmark datasets to achieve high type accuracy, precision, and recall for diverse scented records where additional information may be artificially/mechanically found. The centroid SOM research of olfaction involves investigating more physiologically and nutritionally feasible methods for mapping, understanding, and interpreting massive scent datasets for real-world applications.
Results and conclusion: In all analyzed result and conclusion, the accuracy, precision, and recall of the clustering centroid with optimized feedback SOM are superior to the existing clustering approach. By simulating the data on different set of test and train data it has observed that Proposed (Cluster Centroid with SOM ) method is effective than the existing (Centroid) method. For example, 10% of test data existing method has 67.55% of accuracy and proposed method has 86.75% which is shown in result and conclusion section in details.
Research implications: The research makes an effective contribution by demonstrating the potential and the need to adopt sustainable practices in the management of contemporary companies.
Originality/value: The results and conclusion obtained in this research are unprecedented, innovative and relevant to the medico health community to avoid health diseases, in the context of reliability in social community suggest eating a fresh and pleasant food to avoid health diseases.
Keywords: Centroid Clustering with SOM (CCSO), Electronic Nose (E-Nose), Graphical Neural Network (GNN), Machine Learning, Nutrition, Odor and Self Organizing Map (SOM).
RESUMO
Propósito: O objetivo desta pesquisa é realizar um método inovador baseado em inteligência de aprendizagem automática para determinar a qualidade dos alimentos que podem ser perigosos para a saúde se consumidos por seres humanos. Este artigo detecta a doença humana ao perceber a nutrição que provoca cheiros.
Estrutura teórica: Nos países em desenvolvimento, as pessoas apenas se concentram nas necessidades básicas de alimentos, em vez de se concentrarem na qualidade e nos valores nutricionais dos alimentos que estão exibindo o impacto perigoso de alimentos não saudáveis na vida das pessoas. Muitas pessoas estão sofrendo de diabéticos, cânceres, problemas cardíacos, problemas de fígado e problemas de saúde relacionados ao estômago que são originados devido ao consumo de alimentos ruins. Os consumidores estão satisfeitos com a qualidade dos alimentos, e mais indivíduos estão avaliando-a.
Método/projeto/abordagem: Como metodologia, um nariz eletrônico usa sensores químicos para identificar odores complicados. Tecnologias padrão podem detectar gases de residências, indústrias e materiais explosivos. Não pode cumprir requisitos de frescura. Os narizes eletrônicos, a visão computacional e outras abordagens sensoriais podem imitar as qualidades olfativas, gustativas, visuais e sensoriais humanas, tanto agradavelmente como desagradavelmente. As redes neurais organizam sistemas inovadores de inteligência artificial/mecânica para interpretar gravações de fragrâncias para o reconhecimento do cérebro humano. Inspirados pelo processamento do cérebro humano, oferecemos feedback otimizado, organização por clusters de centroides e mapas auto-organizados para sistemas de aprendizagem de máquina para identificar dados de cheiro. Este trabalho propõe uma técnica de simulação baseada em conjuntos de dados de referência para alcançar alta precisão de tipo e recordação para diversos registros perfumados onde informações adicionais podem ser encontradas artificialmente/mecanicamente. A pesquisa de olfacção no centroide SOM envolve investigar métodos mais fisiológica e nutricionalmente viáveis para mapeamento, compreensão e interpretação de enormes conjuntos de dados de aromas para aplicações do mundo real.
Resultados e conclusão: em todos os resultados e conclusões analisados, a precisão, precisão e recuperação do centroide de aglomeração com SOM de feedback otimizado são superiores à abordagem de aglomeração existente. Ao simular os dados em diferentes conjuntos de dados de teste e treinamento, observou-se que o método proposto (Cluster Centroid com SOM ) é eficaz do que o método existente (Centroid). Por exemplo, 10% dos dados de teste existentes método tem 67,55% de precisão e método proposto tem 86,75% que é mostrado na seção de resultados e conclusão em detalhes.
Implicações da pesquisa: A pesquisa dá uma contribuição efetiva ao demonstrar o potencial e a necessidade de adotar práticas sustentáveis na gestão de empresas contemporâneas.
Originalidade/valor: Os resultados e conclusões obtidos nesta pesquisa são inéditos, inovadores e relevantes para a comunidade médica de saúde para evitar doenças de saúde, no contexto da confiabilidade na comunidade social sugerem comer um alimento fresco e agradável para evitar doenças de saúde.
Palavras-chave: Agrupamento Centroide com SOM (CCSO), Nariz Eletrónico (E-Nose), Rede Neural Gráfica (GNN), Aprendizagem de Máquinas, Nutriçao, Odor e Mapa AutoOrganizado (SOM)
(ProQuest: ... denotes formula omitted.)
1INTRODUCTION
Olfaction is the sense of smell that humans and other living beings share. Humans ought to be able to smell and taste things and activate an excellent memory, i.e., a pleasurable or terrible recall, after some time (Patil Dipti, 2021). This topic aims to teach you about olfactory identification, which is defined as an individual's intake of nutritious items as well as inhaling or exhaling gas. Humans do not recognize the gas with their necked eyes, but sensory neurons allow them to distinguish it 50 times better [(Xiaohui Weng, 2020),(Garner CE, 2009 Nov),(Andres Gongora, 2018),(Vanarse, Espinosa-Ramos, Osseiran, Rassau, & Kasabov, 2020),(Wilson, 2009),(Wilson, 2009)]. We plan to propose designing an artificial olfaction system for health analysis via intelligence learning techniques due to humans' limitations in detecting smell. We motivate the Google blog to train artificial intelligence to predict smell molecules. Research scholars try to develop the neural network and create an artificial flavor from the molecules, e.g., vanilla. Considering the neural network, the sensory neurons can be transmitted through the different layers and updated with the help of graphical neural networks (GNN). A GNN is used to reduce the graph to a vector (the sum of nodes) by forwarding the prediction of neural networks. Here, we learned how to make a database for analyzing different smells related to health. The database for different gases was found on the internet (http://mrpt.org/robotics datasets) for this review. In the experiment of an electronic nose, a total of six electrochemical sensory arrays that found the smell for analytical purposes were used (Garner CE, 2009 Nov). Three are commercial beverages, while three are polish removers based on acetone, regular ethanol, and lighter gas, usually butane but occasionally blended with propane [(M., 1985),(Vanarse, Espinosa-Ramos, Osseiran, Rassau, & Kasabov, 2020)]. TGS2600, TGS-2611, TGS-2620, MICS-5135, and MICS-5521 are the six array sensors utilized to identify the exhaled breath olfaction.
Nutrition is about studying the nutrients in foods, how the body uses them, and the relationship between diet, health, and illness. Nutrients molecular employ ideas from biology, biochemistry, and genetics to comprehend how nutrients affect the human body. In humans, the sense of smell can be crucial. It is constantly felt, whether it is pleasant or unpleasant. The aroma of a person's body can significantly impact their attractiveness. According to a study published by the University of California, everyone has an olfactory signature that is affected by their health, genes, and hygiene diet. Sulfur is usually present in food, which is why broccoli, cauliflower, garlic, and other vegetables produce a pungent body odor. Perception of organic toddler food may affect the human health because of their chemical fertilizers and pesticides. Consumers are more conscious to eat healthy and safety food which provide good nutrition to organic toddler food via olfaction method which determine good human health (Radhika Priya K. P, 2023). Good agriculture community innovate a new technology with digital tools were used to track the growth of crop, weather condition as well as nutritional facts in concern food. Coffee beans and banana are dominant crop as compare to other, such as mango and citrus food found in coffee capital of the Philippines (Corpuz, 2023).
In poorer nations, low-quality, nutrient-poor food may harm health. Bad eating in impoverished nations causes several major health issues:
Bad food quality causes malnutrition, a major health issue in underdeveloped nations. Malnutrition may restrict development, lower immunity, and raise illness risk. Micronutrient deficiencies: poor diet lacks vitamins and minerals. Iron deficiency causes anemia, whereas vitamin A deficiency causes blindness. Obesity, diabetes, and cardiovascular disease are caused by poor diets. These illnesses are rising in emerging countries and may have serious health and economic effects. Foodborne sickness: Poor food may cause serious foodborne illness. Bacterial, viral, or parasite foodborne disease may induce diarrhea, dehydration, and death. Pesticide and fertilizer pollution: Bad food production pollutes the ecosystem. Environmental pollution may harm people and animals.
Improper eating in underdeveloped countries has serious health effects that need collective action. This involves increasing access to nutritious, cheap food, supporting healthy eating habits, and adopting food safety rules to assure food safety and quality. Addressing the social, economic, and environmental causes of substandard food is equally crucial to improve the health and well-being of emerging countries. Use of machine intelligence to determine quality of food for maintaining good health: Machine learning (ML) and Artificial intelligence (AI) can identify food quality. Machine intelligence may aid:
* Quality control: big food quality datasets can educate ML algorithms to identify patterns and attributes that signal food quality. Taste, texture, appearance, and nutrition are examples.
* Food safety: AI can analyze product testing, social media, and news headlines to discover food safety issues including microbial contamination and chemical residues.
* Supply chain management: AI can monitor food from farm to people health shown in table1. These may detect quality concerns and ensure food is kept and delivered properly.
* Predictive modeling: ML learning algorithms can forecast food quality based on its production, storage, and surroundings. This may help you identify and address quality issues before they occur.
* Customized nutrition: AI can analyze genes and dietary preferences to create personalized nutrition recommendations that increase health and food quality. Machine intelligence might drastically alter food quality evaluations. Data and algorithms may make food safer, enhance supply chain management, and create healthy, personalized nutrition regimens.
Do old folks constantly say, an apple a day keeps the doctor away? It turns out to be true that an apple a day keeps bad breath away. Polyphones are antimicrobial substances found in apples (Johnson, 1992). Apples contain a strong enzyme aroma that may help them fight bad breath. It includes high fiber content, necessitating additional chewing time, and increasing saliva production (Kershaw). A well-hydrated mouth must maintain a healthy balance and a higher consumption of fruits and vegetables, and balanced health is less likely to result in health problems, including bad breath. Caries disease is more common when apples are consumed at meal time. Freshness of fruits & vegetables are important to keep good health.
Another reason to eat apples with your dinner is that they can help counteract the stench of other vegetables and fruits. For example, when eating an apple, you can determine its excellent quality of nutrition by considering the odor, i.e., fresh, less fresh, or not fresh, because apples have natural enzymes that help to break down the sulphur compounds found in onions.
Trimethylaminuria, or fishy odor syndrome, is a condition. This is an uncommon ailment that gives off a repulsive fishy smell and makes individuals smell bad. From person to person, this smell is different. To prevent this, we must use our sense of smell to evaluate the nutritional value and quality of the fish. So, wearable technology like an electronic nose that can be used to check foods like fish, meat, and beef must be made so that people don't get sick from eating them.
Objective of this research offers:
a) An innovative approach based on machine intelligence to evaluate the quality of potentially dangerous to human health foods.
b) Using ML algorithms, the nutritional composition of food items may be analyzed.
c) In order to attain high levels of type accuracy, precision, and recall for a wide variety of scented records where extra information may be identified and won automatically.
d) This study provides a simulation approach based on benchmark datasets.
2THEORETICAL FRAMEWORK REFERENCES
The freshness of meat, beef, fish, fruits, and vegetables is an essential indicator of product quality and safety. Today, people live very different lives, the food processing industry is more competitive than ever, and hypermarkets offer a wide range of food options. Manufacturers and distributors want to make sure that the products they sell are of high quality so that their customers don't get sick. Sensory evaluation uses a person's senses to find out about a product's color, smell, and other things.
2.1 Inheritances of Body Odors
2.1.1 Body Odor
Body odor, a widespread issue and an unpleasant natural stench that mostly comes from human bodies, is a concern. People consume food to maintain their health and stave off disease. When consuming nutritious items, keep in mind that they contain both fresh and not-so-fresh features, i.e., rot, sweetness, pungent flavor, mustiness, etc. The body gradually exudes hundreds of flavors and mixtures of volatile organic compounds (VOCs) through metabolites from different somatic cells. Figure 1 illustrates how VOCs are derived from the primary sources of breathing, perspiration, skin, urine, and the face, as well as the transmission of body odor.
2.1.2Breath
Breath contains thousands of volatile organic compounds that have been attributed to either exogenous or endogenous volatile gases. Exogenous gases encompass inhaled organic compounds from the exterior environment that could be produced from oral ingestion of meals or products and compounds from smoking.
There are several other types of endogenous gases, including volatile substances found in the blood that are expelled into the air and metaphorical bacteria that are expelled from the lungs. Nano-molar to pico-molar levels of volatile organic compound concentrations in exhaled air is discovered (Nilakshi Maruti Mule, 2021). Hence, it is always challenging to tell the difference between VOCs generated inside and those from the outside. Yet, taking a breath sample requires no special equipment and causes no discomfort. As a consequence, several GCMS studies of breath samples were conducted, with some resulting in the identification of VOCs unique to a certain condition by academics and doctors.
For example, trimethylamine has been found in the breathing of patients with trimethylaminuria. Acetone was found in the breathing of diabetic patients, and methyl mercaptan was found in the breathing of sweet-smelling patients.
* Skin and sweat: Pores and the skin surface release volatile organic compounds (VOCs), with the most common sources being sweat (a fluid generated by moisture glands) and sebum (an oleaginous liquid secreted by sebaceous glands situated in the axilla, perianal area, and areola of the breasts). Many volatile organic compound (VOC) impacts seem to be due to representative bacteria that reside on the skin's surface, metabolizing and transmuting released chemicals in moisture and sebum. Variations in both the concentration and concentration rate of volatile organic compounds (VOCs) may be a symptom of bacterial infection or an inherent metabolic dysfunction of the sick region. Lacerations caused by cancer or other contagious diseases can leave distinctive marks and smells on the victim. The odor component released by the afflicted region may be collected directly onto an absorbent stable-phase micro-extraction thread or by wiping the condition onto the skin with a natural solvent combined with acetone.
2.1.3Urine
When analyzing urine for illness diagnosis, it is possible to characterize profiles of individual components (others A. Z., 1981). Many metabolic processes use the chemical, and it is ultimately excreted in urine as a byproduct or intermediary. The ketone, alcohol, furan, and sulphide molecules included in this compound are frequently responsible for the drug's characteristic odor. Patients with cancer may have a different pattern of volatile organic compounds (VOCs) in their urine compared to a healthy control; however this variation is disease specific. As a result, urine is made up of a wide variety of substances. Chromatographic data must be extensively processed by a computer in order to identify VOC patterns. And then there were also big disparities in the volatiles profile of people's urine. Interestingly, the metabolic states of the body and the food and drink consumed have significant effects on the components of urine. Those who have just consumed asparagus, for instance, may notice a sulphurous odor in their urine. Hence, caution is needed when assessing if a prospective VOC biomarker is attributable to disease-related metabolic alterations (Pelchat, 2011).
2.1.3.1Particular Disease Odors
Chemical characteristics and the generation of odor mechanism are discussed for various illness scents in Table 1. Infectious illnesses, metabolic abnormalities, poisons or poisoning, and so forth are all triggered by chemical molecules (M., 1985).
* Cholera: Diseases cholera is caused by a pathogenic bacterial infection called Vibrio cholera. This infection causes generous symptoms like bad watery diarrhea, vomiting, and rapid dehydration of the human body system. The behavior of patients with this disease is mentioned as 'rice-water stools' and has pretty sweet odor characteristics (Garner CE, 2009 Nov). This disease is usually spread through contaminated water.
* Advanced breast cancer: Patients with advanced breast cancer often complain of unpleasant odors due to fungal damage, a disease known as "burdens" or "ulcerative lesions," which is characterized by ulceration and proliferation and occurs when malignant tumors cells infiltrate and erode through the pores of the skin. Ulcers are often caused by bacteria, and the affected area frequently stinks [(Wilson, 2009),(others M. S., 2009),(Swanson, 2020),(M., 1985)].
* Gynecological tumors: Some patients with gynecological malignancies may have profuse vaginal discharge with a foul odor because of the emission of acetic acid, isovaleric acid, and/or butyric acid. One potent antibiotic for anaerobes, metronidazole, has been demonstrated to reduce emissions and odors caused by these tumors. If diseasespecific odors are found, they may be used to help cancer patients cope with their condition.
* Small pox: The Variola virus contaminates the body and causes smallpox. Patients who have this illness complain of a high temperature, low back discomfort, nausea, and a rash that is covered with pus. Due to its unpleasant nature, the patient's body exudes a rather sweet and strong smell (M., 1985).
* Pneumonia: It is a respiratory illness caused by a bacterial, viral, fungal, or parasitic infection of the lungs. Alveolar inflammation, fluid-filled lungs, a cough, a fever, and respiratory distress are all symptoms of this illness. Those with pneumonia may have a particularly unpleasant odor coming from their mouths.
* Tuberculosis: Mycobacterium tuberculosis causes a bacterial infection that manifests as TB. While the lungs are the most typical target, this kind of bacterium may affect any area of the body (Karami, 2020). Tuberculosis, which may be detected by a putrid odor while breathing, can damage several systems in the body. A localized swelling in the neck is a rare but possible symptom of TB. Lymph ulcers cause a musty beer odor in certain people.
3METHODOLOGY
3.1Data Processing Technique: Gas Sensory Array
Human olfactory research consists of several effective methodologies, as shown in Figure 2. An electronic nose's hardware and raw input digital records consist of a gas sensor array, a smell subsystem that transports the smell to a digital instrumentation device, and a laptop for record acquisition.
Figure 2 is a block diagram of digital machine olfaction, depicting a set of sensors that can detect (Benabdellah, 2017), identify, and measure volatile organic molecules using artificially optimized feedback mechanisms. The array sensor is considered in the hardware section, while the pattern analysis technique used to process the raw data for the detection and decrease the original data to identify gas molecules is discussed in the following four sections. The goal of dimensionality reduction is to simplify and clarify a dataset by reducing the number of variables in it.
With classification, regression, or clustering, you can use the feature vector of the dimensionality reduction vector to predict how volatile a molecule is. The regression technique is used to identify the learned odors in unknown samples. This method aims to predict the concentration and quality of a typical complex mixture of gases. Finally, the clustering technique is used to learn how the molecular structures of a mix of odors are related. Validation techniques are used to avoid the error rates in an artificial learning model by providing optimized feedback to each method for volatile molecule validation.
Figure 3 (a) shows a graph of the raw odor data for graphic hierarchical clustering. Six MOS array sensors were used to investigate hierarchical clustering responses with different beverages (Wilson, 2009). The dendrogram shows suitable identification of alcohols, especially by the normalized sensor array shown in Figure 3 (b). Self-assembled clustering centroids were mapped to analyze 36325 beverages and liquors from 9 regions based on six fatty acids [(others J. L., 2019),(F Winquist, 1993)]. The clustering centroid SOM can distinguish between nine different areas with well-defined boundaries for the various odors of the inactive molecules. Additional data processing applications (Vanarse, Espinosa-Ramos, Osseiran, Rassau, & Kasabov, 2020) include hierarchical clustering and SOM (Gutierrez-Osuna, 2002) clustering techniques.
To promote record validation, many pattern-based tactics may be used. For each power estimation block, the validated artificial electron nose's intended output offers optimum feedback (F Winquist, 1993). When fresh data analysts are trying to figure out how to get the desired odor to work at its best, they apply a variety of processing techniques. Instead of evaluating the correlation of training datasets to measure performance, one could instead use the capacity of related models to predict novel fundamental data structures.
The sample phases were called Baseline 0, Transient Exposure 1 (which included transition and steady state), and Recovery 2. Figure 4 shows how a metal oxide semiconductor (MOX) gas sensor works when alcohol is present to give a short input pulse signal. This is an ideal model response to stimuli, as an experimental increase in temporary exposure to recovery is envisioned [(Villarreal, 2016),(Restrepo, 1993),(Kermit, 2002)]. Recovery time compared to baseline (start time) and volatile exposure time, i.e., the peak of the maximum intensity of odor concentration, is too long. As shown in Figure 5, scent as a pulse is abnormal and persistent in most cases, recognizing various changes in concentration while navigating without a proper ventilation system. The signal is characterized by continuous fluctuations generated by a unique sensory mechanism.
The combination of gases with a greater intensity in a closed ventilation system is seen in Figure 5. The mixture of gases' concentration, which takes a while to detect via applications, is classified as baseline, volatile exposure, and recovery. The initial behavior is thought to be the best starting point for a gas because of the raw vacuum, but with time the strength of the odor may develop owing to inadequate ventilation, as seen in Figure 4. To reduce the strength of the gases in the environment and restore it to its normal condition, the gases must flow in a 3600-degree direction while being mixed with oxygen [(Bax, 2020),(Kermit, 2002),(Seto, 2006),(Moshayedi, September 2013)].
3.2 Data Analysis
The pattern recognition algorithm used to analyze the data produced by the sensor is crucial to the simulation of the electronic nose. Figures 4 and 5 show the examination of the data needed for the nonlinear and adaptive curves. Because sensor array performance varies by superimposing the intricate details of odor detection, the data analysis methods used are critical. Data preparation, dimensionality reduction, classification/regression/clustering and validation are the four components of the data analysis step.
3.3 Pre-Processing
The odor of the human body, as well as the body and foreign substances in the body, generate gas. The sensor array collects these volatiles. The sensor array detects the strength of the odor and records it in the dataset. The number of features is extracted and provided to the dimensionality reduction procedure, while the raw dataset record is given to the pre-processing phase. Pre-processed data may be used to map odor intensity levels and remove unnecessary fields from recordings. Equations 1 and 2 show the extraction measure and extraction matrix, respectively.
VEi = (Ei1Ei2...Ei0) (1)
... (2)
'o' is the number of E-Nose odor variables and 'n' is the number of samples.
A digital e-nose with a variety of gas equipment has been chosen to record smells that match up with how fresh a product is. The digital feature variable of the ith sample is created by extracting the best response from each sensor (1). To figure out what the extraction matrix is like, all of the samples are taken frequently (2).
In response to the odor presentation, the array of gas sensors sends a signal to x that depends on the sampling time ('t') and the number of sensors ('n') used. The gas sensor array's response to a single sample measurement is xn(t). Where n = 1, 2, .. ..N and 'N' is the number of sensors present in the array. In order to research the biological sense of scents and identify odors across time, classification, regression, and clustering in practice all need the signal xn(t). This will enable the E-Nose gadget to carry out comparable odor detection capabilities.
3.4Dimensionality Reduction
Each gas sensor reacts to the peculiarities of the odor. Following reduction, each gas sensor's response to a single feature is extracted. The vector Veí = (Ei1 Ei2.Eio) t, whose dimension is equal to the number of sensors in the array, represents the response of the whole gas array sensor. These 'o' odor dimensions have overlapping sensitivity and considerable redundancy. As a result, certain pertinent predictions may be used for visualization as well as reduction.
Principal component analysis is often used to simplify large amounts of data (PCA). The principal component analysis is a method that maximizes the variation around the data that was previously classified. After an assessment of the Eigen values, the PCA output is converted to data, and then only the Eigen values are consciously considered in the subsequent analysis. Hence, the literature details the steps necessary to streamline the information collected by a gas sensor.
4CLASSIFICATION/REGRESSION/CLUSTERING
4.1 Classification
The classification process depends on the ability to recognize unprocessed smells. Several techniques were taken into consideration to categorize unidentified odor measurements or to separate unknown odors from a predefined dataset of expected odors. The k-means algorithm was probably the most popular technique for unsupervised classification. It has also been noted that self-organizing Maps (SOMs) are used. When commercial software became available, artificial neural networks were utilized for supervised classification.
4.2Regression
Regression problems produce regions where e-nose application use is more challenging. You must build a regression prediction model for a set of independent variables and a set of continuous dependent variables. Although the established variables are regarded as an explicit problem, the sample pattern classification can be handled as a regression problem. This enables you to carry out a regression strategy for the category's root cause.
4.3Clustering
Techniques for unsupervised learning, like clustering, are used to find high-dimensional features and show correlations or similarities between data samples in three dimensions. Finding "appropriate" examples of clusters involves (i) finding dissimilarity (typically measured by Euclidean distance), (ii) using optimized clustering (typically based on topologically ordered clusters (Cuevas Rivera, 2015), and (iii) using centroid clustering, includes a definition of the search algorithm. In contrast to the monitored process, validation typically calls for a domain expert, and the outcomes are both subjective and objectively measurable.
4.4 Validation
Validation occurs when the model is used to make predictions about untested data. If you want your model to function at its best, it's important to choose the right one for your data analysis. A dataset with an excessive number of epochs may occur if the model is over fit. Each dataset is split into a training set and a validation set to avoid over fitting. Both the standard model and the training model are used to learn different types of learning structures, and the standard model is then utilized for authentication to determine which model is selected. Results in the final model may be enhanced by using a centroid / clustering SOM to get higher precision and precision search parameters.
4.5 Proposed Centroid SOM
Unsupervised learning techniques can produce centroid topological clusters of artificial neurons. SOM is applied to the high-dimensional datasets. Two-dimensional self-organizing topological centroid clustering is used to create the dendrograms, which are geologically stored as long or short distances on the map. The SOM map consists of a position path vector (1) on a 2D grid that carries the weights of the same molecule based on the measurement of the input information molecule. First, in the mastering part of the SOM, a random weight is derived from the gas depth, then an epoch is performed, and finally, the centroid grid distribution fact vector captures the gas information. This works most closely with the validation output. To obtain the provided dimensionality reduction matrix of the centre of gravity cluster, the distance between gas molecules in the excess dimensional topology is brought to the 2D SOM projection grid as 3D. The SOM projection grid results from the completion, cooperation, and adaptation processes. Gas molecules (neurons) are captured by sensors and participate in the competition. After feature extraction and activation, the closest molecule is selected as the winner. Finally, all activated gas molecules approach the input pattern and take feature coordinates to identify and validate the ML output using data visualization through the exciting properties of the SOM.
Figure 6 depicts the procedures required to analyze the nutritional odoration diagram in Figure 2 using the building blocks of pattern analysis. Six gas sensors were arranged in a sensor array, gathering combined raw data. After processing the application that it has in order to predict the odor of the hierarchical group, the class label for each odor in the feature vector is added. The dendrogram of a discrete set of learned and known labels for hierarchical clustering is used. To increase posterior probabilities, the classification error should be reduced if the model detects an error. Bayes' theorem should be used to cut out pointless label sets and improve calculations for odor performance in order to reduce this error.
Different techniques were used to reduce the matrix for test data as well as training data to validate the volatile organic compound volatile molecule of odor. Later it has to calculate the transpose vector by applying SOM model to test odor matches found with similarity type i.e. 1, 2, 3, 4, 5 and 6.
5RESULTS ANALYSIS AND DISCUSSION (EXPERIMENTATIONS)
5.1 Data Description and Performance Criteria
Experiments have made use of six different gas detectors, such as the TGS-2600, TGS2602, TGS-2611, and TGS-2620, as well as the MICS-5135 and MICS-5521. The application of this experiment is used in households, in industries to detect explosion gases, in kitchen communities to avoid the accident of cylinder explosions as well as to eat fresh food products, and it is also used to find the person who is drunk and driving. All of these applications work in tandem with one another.
The raw data set has been taken from http://mrpt.org/robotics_datasets. Figure 7 depicts the top ten rows of the raw dataset.
The six sensors in the TGS-2600, TGS-2602, TGS-2611, TGS-2620, MICS-5135, and MICS-5521 sensor arrays are responsible for gathering the readings for the columns H, G, I, J, F, and E, shown in figure 7 & 8 respectively. The name of the gas is shown in columns A, B, and C along with a time stamp and its temperature value. The remaining fields of a dataset may be used as-is without further processing. There are 36325 rows and 35 columns in the raw datasets. The datasets are reduced to a new shape with 36325x8 rows and columns by removing the unnecessary columns. Figure 9 depicts the cluster centers and their corresponding cluster centroid SOM sensed data. The results are calculated and taking into an account of accuracy, precision, and recall.
To improve upon the outcome presented in Table 2, we must provide centroid clustering with optimal feedback through a self-organizing map (SOM). Accuracy, precision, and recall are compared in Table 2 for the current clustering strategy, the recommended centroid clustering with SOM (CCSOM).
The accuracy, precision, and recall of the clustering centroid with optimized feedback SOM are superior to the existing clustering approach, which is shown in Figure 10. We have tested accuracy on 10, 15, 20, 25, and 30 percent of the test data sets for both the existing and proposed methods (Table 3).
By simulating the data on different set of Test and train data it has observed that Proposed (Cluster Centroid with SOM) method is effective than the existing (Centroid) method. On 10% of test data existing method has 67.55% of accuracy and proposed method has 86.75%. While considering 15, 20, 25 and 30 % test data existing method has around 67.5% of accuracy, 58.8% of Precision and 58.7% of Recall. On other hand proposed system has around 87% of accuracy, 68% of Precision and 68% recall. Table 7 shows detail Comparison of Accuracy, Precision, Recall and False positive of existing (Centroid (%)) and Proposed CCSOM (Cluster Centroid with SOM (%)).
Figure 9 to 13 shows the comparison of Accuracy between Existing (Centroid (%)) and Proposed System (Cluster Centroid with SOM (%)). By the graph it is clearly visible that proposed system is more effective than the existing system for all different set of test data. Table 7 gives detail comparison between existing vs. proposed method. Proposed CCSOM method is more effective than that of existing method and this method can be used for Assessing Nutritional Olfactory Features for Health Analysis.
6CONCLUSION
Several foods and nutrients are essential to human health. Human body scents might be represented by higher dimensional datasets. The olfactory function of a human-like smell may be simulated using traditional ways for the aim of judging if a manner is pleasing or repulsive. The results of this investigation suggest that the human sense of smell may have a role in the detection of sickness during certain meals. Diseases may be diagnosed via the use of olfactory cues by using ML techniques (inhalation or exhalation). As a result of the self-organization feedback, the E-nose device's accuracy has been increased, which in turn has informed the appropriate intensity level of gas levels. Researchers found that their suggested CCSOM method outperformed the status quo in terms of accuracy, precision, and recall. Further work will focus on honing the algorithm's precision.
The nutrients in meals are essential to maintaining good health since they fuel the body's systems. To a large extent, our health is affected by the nutritional value of the food we consume. Certain health issues, such obesity, type 2 diabetes, heart disease, and some malignancies, have been linked to eating poorly, such as highly processed or fast food. The quality of food may be evaluated in a number of ways with the help of ML. More informed customers may make healthier food choices with the assistance of machine intelligence employing these approaches. Using ML algorithms, the nutritional composition of food items may be analyzed. This may assist discover vitamin deficits and excesses, both of which can have negative effects on health.
There is also the possibility of developing an app and a gadget that doctors may use to diagnose illnesses only by smelling the patient's breathe. Dietary supplement and illness prevention suggestions will be provided as part of this research.
REFERENCES
Andres Gongora, J. M.-J. (2018). An Electronic Architecture for Multipurpose Artificial Noses. Journal of Sensors, vol. 2018, Article ID 5427693, 9 pages. doi: https://doi .org/10.1155/2018/5427693
Bax, C. S. (2020). How Can Odors Be Measured? An Overview of Methods and Their Applications. Atmosphere, 11(1), 92. doi:https://doi.org/10.3390/atmos11010092
Benabdellah, N. &. (2017). Identification of two types of rotten meat using an electronic nose for food quality control. International Journal on Smart Sensing and Intelligent Systems, vol. 10, no. 3, 673-695. doi:10.21307/ijssis-2017-229
Corpuz, J. T. (2023, 06 05). Shifting Paradigms in Coffee-Based Cultivation: Cultural Management Practices and Technology Utilization in Upland, Cavite, Philippines. International Journal of Professional Business Review, Vol 8(No 6), e01793. doi:https://doi.org/10.26668/businessreview/2023.v8i6.1793
Cuevas Rivera, D. &. (2015). Modelling Odor Decoding in the Antennal Lobe by Combining Sequential Firing Rate Models with Bayesian Inference. PLoS computational biology., vol. 11(no. 10), e1004528. doi:10.1371/journal.pcbi.1004528
Dankert J, H. Y. (1981, Dec 5). Metronidazole in smelly gynaecological tumours. Lancet. 2(8258):1295. doi:doi: 10.1016/s0140-6736(81)91539-7. PMID: 6118711
F Winquist, E. G. (1993, Dec.). Performance of an electronic nose for quality estimation of ground meat. Measurement Science and Technology, Volume 4(Number 12), 1493-1500. doi:DOI 10.1088/0957-0233/4/12/029
Garner CE, S. S. (2009 Nov). A pilot study of faecal volatile organic compounds in faeces from cholera patients in Bangladesh to determine their utility in disease diagnosis. (E. 2. 19268999, Ed.) Trans R Soc Trop MedHyg, 103(11):1171-3. doi:doi: 10.1016/j.trstmh.2009.02.004
Gutierrez-Osuna, R. (2002, July). Pattern Analysis for Machine Olfaction: A Review. IEEE Sensors Journal, vol. 2(no. 3), 189 - 202. doi:10.1109/JSEN.2002.800688
Inbar Nardi-Agmon, M. A.-H.-M. (2016, June). Exhaled Breath Analysis for Monitoring Response to Treatment in Advanced Lung Cancer. Journal of Thoracic Oncology,, Volume 11(Issue 6). doi:https://doi.org/10.1016/jjtho.2016.02.017.
Johnson, B. (1992, Nov). Halitosis, or the meaning of bad breath. Journal of General Internal Medicine, vol. 7(no. 6), 7, 649-656. doi:DOI: 10.1007/BF02599209
Karami, H. &.-G. (2020, October). Qualitative analysis of edible oil oxidation using an olfactory machine. Journal of Food Measurement and Characterization, vol. 14(no. 5), 26002610. doi:10.1007/s11694-020-00506-0
Kermit, M. &. (2002, September). Intelligent Machine Olfaction. Conference: IASTED International Conference Artificial and Computational Intelligence. Tokyo, Japan.
Kershaw, J. C. (n.d.). Nutrition and taste and smell dysfunction. World journal of otorhinolaryngology - head and neck surgery, vol. 4(no. 1), 3-10. doi:https://doi.org/10.1016/j.wjorl.2018.02.006
M., T. (1985). Foul breath and a productive cough. Hospital practice (Office ed.), , 20(5A), 50.
Maruyama, X. a. (2002). The Once and Future Threat of Smallpox," Naval War College Review. Volume 55(No 2 , Article 11). Retrieved from https://digital-commons.usnwc.edu/nwcreview/vol55/iss2/11
Moshayedi, A. T. (September 2013). Mathematical modeling for SnO2 gas sensor based on second-order response. 2013 IEEE Symposium on Industrial Electronics & Applications (ISIEA), (pp. 33-38). doi:DOI:10.1109/ISIEA.2013.6738963
Nilakshi Maruti Mule, D. D. (2021, August 31). A comprehensive survey on investigation techniques of exhaled breath (EB) for diagnosis of diseases in human body. Volume 26. doi:https://doi.org/10.1016/j.imu.2021.100715
others, A. Z. (1981, June 1). The role of organic volatile profiles in clinical diagnosis. Clinical Chemistry, Volume 27(Issue 6), Pages 789-797. doi:https://doi.org/10.1093/clinchem/27.6.789
others, J. L. (2019, January ). Machine Learning in Human Olfactory Research. Chemical Senses, Volume 44(Issue 1), Pages 11-22. doi:https://doi.org/10.1093/chemse/bjy067
others, M. S. (2009, September 23). Dimethyl Trisulfide as a Characteristic Odor Associated with Fungating Cancer Wounds. Bioscience, Biotechnology, and Biochemistry, Volume 73( Issue 9), Pages 2117-2120. doi:https://doi.org/10.1271/bbb.90229
Patil Dipti, M. D. (2021, March ). Digital Representation Techniques For Olfactory Features. Information Technology In Industry, vol. 9 (No. 1). doi:10.17762/itii .v9i1.131
Pelchat, M. L. (2011). Excretion and perception of a characteristic odor in urine after asparagus ingestion: a psychophysical and genetic study. Chemical senses, 36(1), 9-17. doi:https://doi.org/10.1093/chemse/bjq081
Ponticorvo, M. &. (2020). Applied Behavior Analysis (ABA) as a Footprint for Tutoring Systems: A Model of ABA Approach Applied to Olfactory Learning. Social Sciences. doi:10.3390/socsci9040045
Radhika Priya K. P, D. A. (2023, 06 07). Organic Toddler Food - an Analysis of the Perception of Users. Revista De Gestao Social E Ambiental, Vol 17(No 3), e03362. doi:https://doi.org/10.24857/rgsa.v17n3-007
Restrepo, D. O. (1993). Human olfactory neurons respond to odor stimuli with an increase in cytoplasmic Ca2+. Biophysical journal,, vol. 64(no. 6), 1961-1966. doi:https://doi.org/10.1016/S0006-3495(93)81565-0
Seto, S. &. (2006). Mathematical model of semiconductor gas sensor. Sensors and Materials, 18. 001-016. .
Shirasu, M. &. (2011). The scent of disease: volatile organic compounds of the human body related to disease and disorder. Journal of biochemistry, Vol 150(no 3), 257-266. doi:https://doi.org/10.1093/jb/mvr090
Swanson, B. F. (2020). Electronic Nose Analysis of Exhaled Breath Volatiles to Identify Lung Cancer Cases: A Systematic Review. The Journal of the Association of Nurses in AIDS Care : JANAC, 31(1), 71-79. doi:https://doi.org/10.1097/JNC.0000000000000146
Vanarse, A., Espinosa-Ramos, J. I., Osseiran, A., Rassau, A., & Kasabov, N. (2020, May). Application of a Brain-Inspired Spiking Neural Network Architecture to Odor Data Classification. journal-of-animal-physiology-and-animal-nutrition, Sensors, vol. 20, no. 10, 10 p. 2756. doi:https://doi.org/10.3390/s20102756
Villarreal, B. &. (2016). Bioinspired Smell Sensor: Nostril Model and Design. IEEE/ASME Transactions on Mechatronics, 21, 912-921.
Wilson, A. &. (2009). Applications and Advances in Electronic-Nose Technologies. Sensors, 9(7), 5099-5148. doi:https://doi.org/10.3390/s90705099
Xiaohui Weng, X. L.-M. (2020). A Comprehensive Method for Assessing Meat Freshness Using Fusing Electronic Nose, Computer Vision, and Artificial Tactile Technologies. Journal of Sensors, vol. 2020, Article ID 8838535, vol. 2020, 14 pages. doi:https://doi.org/10.1155/2020/8838535
You have requested "on-the-fly" machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Show full disclaimer
Neither ProQuest nor its licensors make any representations or warranties with respect to the translations. The translations are automatically generated "AS IS" and "AS AVAILABLE" and are not retained in our systems. PROQUEST AND ITS LICENSORS SPECIFICALLY DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING WITHOUT LIMITATION, ANY WARRANTIES FOR AVAILABILITY, ACCURACY, TIMELINESS, COMPLETENESS, NON-INFRINGMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Your use of the translations is subject to all use restrictions contained in your Electronic Products License Agreement and by using the translation functionality you agree to forgo any and all claims against ProQuest or its licensors for your use of the translation functionality and any output derived there from. Hide full disclaimer
© 2023. This work is published under https://rgsa.emnuvens.com.br/rgsa/about/editorialPolicies#openAccessPolicy (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Abstract
Propósito: O objetivo desta pesquisa é realizar um método inovador baseado em inteligência de aprendizagem automática para determinar a qualidade dos alimentos que podem ser perigosos para a saúde se consumidos por seres humanos. Este artigo detecta a doença humana ao perceber a nutrição que provoca cheiros. Estrutura teórica: Nos países em desenvolvimento, as pessoas apenas se concentram nas necessidades básicas de alimentos, em vez de se concentrarem na qualidade e nos valores nutricionais dos alimentos que estão exibindo o impacto perigoso de alimentos não saudáveis na vida das pessoas. Muitas pessoas estão sofrendo de diabéticos, cânceres, problemas cardíacos, problemas de fígado e problemas de saúde relacionados ao estômago que são originados devido ao consumo de alimentos ruins. Os consumidores estão satisfeitos com a qualidade dos alimentos, e mais indivíduos estão avaliando-a. Método/projeto/abordagem: Como metodologia, um nariz eletrônico usa sensores químicos para identificar odores complicados. Tecnologias padrão podem detectar gases de residências, indústrias e materiais explosivos. Não pode cumprir requisitos de frescura. Os narizes eletrônicos, a visão computacional e outras abordagens sensoriais podem imitar as qualidades olfativas, gustativas, visuais e sensoriais humanas, tanto agradavelmente como desagradavelmente. As redes neurais organizam sistemas inovadores de inteligência artificial/mecânica para interpretar gravações de fragrâncias para o reconhecimento do cérebro humano. Inspirados pelo processamento do cérebro humano, oferecemos feedback otimizado, organização por clusters de centroides e mapas auto-organizados para sistemas de aprendizagem de máquina para identificar dados de cheiro. Este trabalho propõe uma técnica de simulação baseada em conjuntos de dados de referência para alcançar alta precisão de tipo e recordação para diversos registros perfumados onde informações adicionais podem ser encontradas artificialmente/mecanicamente. A pesquisa de olfacção no centroide SOM envolve investigar métodos mais fisiológica e nutricionalmente viáveis para mapeamento, compreensão e interpretação de enormes conjuntos de dados de aromas para aplicações do mundo real. Resultados e conclusão: em todos os resultados e conclusões analisados, a precisão, precisão e recuperação do centroide de aglomeração com SOM de feedback otimizado são superiores à abordagem de aglomeração existente. Ao simular os dados em diferentes conjuntos de dados de teste e treinamento, observou-se que o método proposto (Cluster Centroid com SOM ) é eficaz do que o método existente (Centroid). Por exemplo, 10% dos dados de teste existentes método tem 67,55% de precisão e método proposto tem 86,75% que é mostrado na seção de resultados e conclusão em detalhes. Implicações da pesquisa: A pesquisa dá uma contribuição efetiva ao demonstrar o potencial e a necessidade de adotar práticas sustentáveis na gestão de empresas contemporâneas. Originalidade/valor: Os resultados e conclusões obtidos nesta pesquisa são inéditos, inovadores e relevantes para a comunidade médica de saúde para evitar doenças de saúde, no contexto da confiabilidade na comunidade social sugerem comer um alimento fresco e agradável para evitar doenças de saúde.