End to End Forest Cover Classification ML Mini Project

• Predict or test model @ https://forestclassclassification.streamlit.app/

Work Flow

1) Data Insertion and Preparation

• Understanding of features with Definitions.
• Creating Pickle files for One hot Encoded and Label Encoded Features.
• Preparing and Exporting the Dataset for Exploratory Data Analysis.

2) Exploratory Data Analysis

• Understanding the data.
• Finding out if the data is imbalanced.
• Outliers detection.
• Skewness and Kurtosis Detection.
• Univariate and Bivariate Analysis
• Making the data features normally distributed.

3) Model training

• Training With different Classification Model
  • Logistic Regression
  • KNN Classifier
  • Decision Tree
  • Random Forest Classifier
  • Balanced Random Forest Classifier
  • Xtreme Gradient Boost Classifier
• Hypertuning Each algo to get the best fit.
• Saving the best model into a pickle file and using for future predictions

4) Saving all pickle files in AWS S3

• This process is done because github file size restriction is 25MB.
• But the Model here was more than 25MB.

5) Streamlit Application

• Reading Pickle Files from AWS S3
• The credentials will not be initialized in streamlit.py file but in streamlit environment for data security.
• Creating Manual Input and also Slider drag drop input for entering feature values
• Inputing data and getting the predictions in the application.

THE END