future-sales-prediction-2024 1.2.4

A package for feature extraction, hyperopt, and validation schemas

Data Science: Sales Prediction

Project Status: Completed

Overview:

This project provides a Python package,future_sales_prediction_2024, to simplify common tasks in data science workflows. It includes tools for feature extraction, validation schema creation, hyperparameter optimization, and model training.

Methods Used

• Feature Engineering: Automates the creation and selection of important features, including memory optimization.
• Validation: Implements schema validation to ensure data consistency, identify missing values, and prevent duplicate records.
• Hyperparameter Tuning: Leverages tools like hyperopt for efficient parameter search.
• Visualization: Includes plotting tools for feature importance and error analysis.

Technologies

• Python
• Pandas, jupyter

Data Sources:

The tools in this package are designed to work with structured datasets, such as CSV files. For example, it can handle datasets used in machine learning competitions like Kaggle or any tabular data source.

Challenges:

• Complexity in Generalization: Making the tools generic enough to work with diverse datasets while maintaining simplicity.
• Performance Optimization: Balancing ease of use with computational efficiency.
• Error Handling: Ensuring clear and helpful error messages for data validation and model failures.

Conclusion:

This package is a modular and flexible solution for streamlining data science workflows. It provides data scientists and ML engineers with reusable tools to focus on solving domain-specific problems.

[0.1.1] - 2024-11-25

Added

• Changes in loader function: upload files using filenames.

[0.2.1] - 2024-11-26

• Added support for Google Cloud Storage.
• Improved deployment pipeline.
• Bug fixes and performance improvements.

[0.2.2] - 2024-11-27

• Bug fixes.

[0.2.3] - 2024-11-28

• Enhanced Explainability and Error Analysis Users can now save plots generated by the Explainability and ErrorAnalysis classes to files. The directory and filenames are customizable, and plots are automatically overwritten if files with the same name already exist.
• Customizable Hyperparameter Tuning Users can now fully customize the hyperparameter tuning process: Define the search space for hyperparameters. Specify the optimization algorithm and objective function. Tailor the evaluation process to their needs.
• FeatureImportanceLayer Enhancements Plots for baseline and final model feature importances can now be saved directly to disk. Customizable output directory (output_dir) and file names. Plots overwrite existing files with the same name.

[0.2.4] - 2024-11-29

• Bug fixes.

[1.2.4] - 2024-11-29

• Cloud Storage Integration
• The data_handling.py and feature_extraction.py scripts now support loading .csv files from GCS paths. Outputs are saved to a user-specified GCS directory via the --outdir parameter.