Machine Learning Model to predict the Type of Meat.
"This is the continuation of research done on Meat Analysis to predict the Meat type on the basis of NIR data."
The project's primary goal is centered around meat data. The purpose is to forecast the correct label of the meat such as Chicken, Turkey, Pork, Beef, or Lamb by reading its related NIR data. In order to achieve the objective, I researched several online resources and books. Finally I used Data Science methodologies such as data exploration, and feature selection/dimension reduction followed by applying Machine learning techniques to get the best Machine Learning model to predict the Meat type.
NIR data was highly correlated and I found the best result by applying Partial Least Squares (PLS) with accuracy more than 96%.
More details for Sklearn.cross_decomposition.PLSRegression
How it works?
Go to Meat App
Step 1 : Upload meats.csv file by browsing the file and hit "Upload Meat Data"
Step 2 : Run the system to create Machine Learning Model. By hitting the "Process Meat Data, system will run the Machine Learning Algorithm to create model. Code will also create Train and Test dataset. Test data set will be used in next step. At the end it will show the accuracy of the Machine Learning model.
Step 3 : To predict the type of a record of NIR data , Hit button "Meat Data" Table will be made available for the user along with the record number as first column. User can enter the record number "Enter Row Number" text field and then hit button predict to see the result. Result will be shown along with success/failure related image and the row used for prediction.
Project Information:
DTSC 691 , Capstone, Applied Data Science
Instructor - Jake Bodet
Mentor - Narayan Devkota
Developer - Abhishek Kumar
Other Machine Learning Algorithm used for the research:
- PCA LOGISTIC , Split Value: 0.23 Accuracy 82.35 %
- PCA TREE , Split Value: 0.3 Accuracy 77.61 %
- PCA SGDC , Split Value: 0.21 Accuracy 76.6 %
- PCA LDA , Split Value: 0.32 Accuracy 81.94 %
- PCA QDA , Split Value: 0.23 Accuracy 84.31 %
- PCA KNN , Split Value: 0.22 Accuracy 87.76 %
- PCA SVM , Split Value: 0.22 Accuracy 83.67 %
- Genetic LOGISTIC , Split Value: 0.2 Accuracy 82.98 %
- Gentic TREE , Split Value: 0.2 Accuracy 74.47 %
- Gentic SGDC , Split Value: 0.2 Accuracy 55.32 %
- Gentic LDA , Split Value: 0.2 Accuracy 78.72 %
- Gentic QDA , Split Value: 0.2 Accuracy 87.22%
- Gentic KNN , Split Value: 0.2 Accuracy 80.85 %
- Gentic SVM , Split Value: 0.2 Accuracy 85.11 %
Bibliography:
Singh, Manokamna, and Katarina Domijan. "Comparison of machine learning models in food authentication studies." 2019 30th Irish Signals and Systems Conference (ISSC). IEEE, 2019.
J. McElhinney, G. Downey and T. Fearn, "Chemometric processing of visible and near infrared reflectance spectra for species identification in selected raw homogenised meats", Journal of Near Infrared Spectroscopy, vol. 7, pp. 145-154, 1999.