classifying the glass type using percentage of minerals present in each class of glass PROJECT TITLE - Glass Classification
GOAL - The aim of this project is to classify the glass type using percentage of minerals present in each class of glass
WHAT HAVE I DONE
Loading datasets Features in data Exploratory Data Analysis and Data Pre-processing Checking distribution of the features Univariate Box Plot Bivariate Box plots Scatter Matrix Pairplot Correlation Plot Feature Engineering - Based on the mean of K and Ca in classes Statistical Importance Check for Variable Splitting the data Model Building and Comparison Analysis Using Logistic Regression Using Random Forest Classifier Using Gradient Boost Using Decision tree Classifier Using KNN Using Support Vector Classifier Saving the Random Forest Classifier model Making classification using the Keras Sequential Model approach Data Pre-processing Building the Keras model Building the Keras model accuracy Model evaluation Saving the Keras model
MODELS USED
Logistic Regression Random Forest Classifier Gradient Boost Decision tree Classifier KNN Support Vector Classifier Keras Sequential Model
LIBRARIES NEEDED
numpy pandas matplotlib scipy seaborn scikit-learn pickle tensorflow
Comparing the models
Logistic Regressor model
Accuracy Score of Logistic Regression : 0.6976744186046512 Random Forest Classifier model
Accuracy Score of Random Forest Classifier : 0.8372093023255814 Gradient Boost classifier model
Accuracy Score of Gradient Boosting Classifier : 0.7906976744186046 Decision Tree Classifier model
Accuracy Score of Decision Tree Classifier : 0.5581395348837209 KNN Classifier model
Accuracy Score of K-Nearest Neighbors Classifier : 0.813953488372093 Support Vector Classifier model
Accuracy Score of Support Vector Classifier : 0.7441860465116279 From the model comparison it is evident that the Random Forest Classifier gives the best results with a score of 83.72%.Now, we are going to build a Keras Sequential model to check whether it performs better than the Random Forest Classifier. Keras Sequential Model Validation Accuracy - 95.24% Testing Accuracy - 90.91%
Conclusion
In this project we have performed a analysis and visualization of the training dataset with different Exploratory Data Analysis techniques. Then a comparative analysis of different ML Classifiers have been done to predict type of glass. After performing the comparative analysis of the classifiers, we can see that the Random Forest Classifier gives the best results with a score of 83.72%. However to get better accuracy we build a Keras Sequential model which gives a validation accuracy of 95.24% and Testing accuracy of 90.91%.