/Binary-Classification-XGB

Assessment solution for Arya-ai Binary classification assessment

Primary LanguageJupyter NotebookMIT LicenseMIT

Binary-Classification-XGB

Assessment solution for Arya-ai Binary classification problem

Project File Struture:

Root /fldr

  • Alpha&Omega.ipynb <---- Main Jupyter notebook (Forgot to change name)
  • Assignement - Data Scientist (1).docx <---- Assessment problem document
  • Testing_predicitons.csv <---- Target class outputs of the Test Data
  • README.md <---- this very file
  • XGBFTW.sav <---- XGBoost model export done over using pickle
  • requirements.txt <---- Environment Screenshot
  • essentials_only_req.txt <---- ipynb specific requirements
  • Data /fldr
    • Training_set.csv <---- Training Dataset
    • Test_set.csv <---- Testing Dataset

Data Stats:

  • Train Dataset Shape -> (3910,58)
  • Test Dataset Shape -> (691,57)
  • Dataset is Sparse and High Dimensional
  • Features are highly skewed

Key Decisions:

  • Used RandomForest Classifier for feature selection.
  • Selected top 30 features with respect to their feature importance.
  • For metric considered Binary CrossEntropy | LogLoss and ROC-AUC score.
  • Model of choice is Xgboost.

Process Flow - Main.ipynb (Alpha&Omega.ipynb)

  1. EDA
  2. Splitting the data
  3. Feature Selection
  4. Data Scaling - Normalization
  5. Model Training
  6. Prediction Metrics
  7. Processing and Predicting on Test Data
  8. Saving Model for Future Usage
  9. Exporting Y_test Predicted scores
  10. Generating requirements. #Has an important Note. Must Read!

Process Flow - Performance_print.py

  1. Splitting the Data
  2. Feature Selection
  3. Importing Presaved model
  4. Using presaved model to generate scores
  5. Using Prettytable to print output table

Had fun making this!!