My Scicrop job application!

Check out the GitHub repository:

This project is a solution to a multi-class supervised learning task. The repository is structured in 5 notebooks in which we explore the data and try different approaches.

1. FirstAnalysis
2. Missing Values
3. Machine Learning
4. Optimize
5. Validate

After we have the best solution we implement them in python scrips, there are 3 main scripts:

1. train_test_split.py
2. create_models.py
3. predict.py

This is a complete data pipeline - from preparing the data, training, optimizing, saving the models and predicting target data. This pipeline can be ran using the make command:

make train_predict

The data pipeline can also be ran using a docker container!

make docker

Here we build the docker image, mount the host results/ directory to the cointainer’s directory and run the pipeline, the prediction output is saved in the host.

📓 First Analysis

• Class Balance
• Missing Data
• Features Analysis
• Clustering

📓 Missing Values

• Missing data
• Train-Validation Subsets
• Data Imputation
• Results

📓 Machine Learning

• Data Cleaning
• Binary Model
• MultiClass Prediction

📓 Optimization

• Binary Classifier
• K fold prediction
• Multiclass Classifier

📓 Validation

• Binary Model
• Multiclass Model
• Validation Data

🖥️ Code Documentation

• lib package
  • Submodules
  • lib.data_processing module
  • lib.model module
  • lib.utils module