How otniel113/mvpcalculator-regresi works
Project type and purpose
• This is a data-analysis/machine-learning pipeline implemented as a Jupyter Notebook (MVPCalculator.ipynb).
• Its purpose is to compute an “MVP score” for Mobile Legends matches using linear regression.
• It ingests manually collected match data, preprocesses it, trains a model, evaluates it, and exposes a simple deployment endpoint.
Main components
a. Data Source
• Google Spreadsheet (manual match history).
b. Data Ingestion & Preprocessing
• Python/Pandas code in the notebook.
• Operations: drop “Gold” & “Minutes”, compute GPM, filter features by Pearson correlation.
c. Feature Selection
• Selection of features: kill, death, assist, gpm, teamfight_part, win, score.
d. Model Training
• Scikit-learn’s LinearRegression.
• Splitting data (80/20 train/test).
e. Model Evaluation
• Metrics: R2, RMSE, MAE.
f. Deployment / Inference
• The final regression formula is hosted as a lightweight service (URL: https://otnielabiezer.com/MVPCalculator/).
• Visual: parallelogram “Web UI / API” where users submit match stats and get MVP score.
g. Environment & Tools
• Jupyter Notebook, Python, Pandas, scikit-learn.
Relationships and data flows
• “Preprocessing” → “Training” (train split).
• “Preprocessing” → “Evaluation” (test split).
• “Training” → “Evaluation” (trained model).
• “Training” → “Deployment” (exported regression formula or model).
• “Web UI/API” → “Model” → “Response with MVP score”.
Architectural patterns and design principles
• ML Pipeline: clear separation between data ingestion, preprocessing, training, evaluation, and deployment.
• Reproducibility: everything lives inside one notebook but could be refactored into modules (.py) following pipeline stages.
• Modular design allows swapping out the regression algorithm or adding more features.
Detailed checklist
[ ] Raw Data (Google Sheet)
[ ] Data Preprocessing (drop, feature engineering)
[ ] Feature Selection (Pearson correlation)
[ ] Model Training (LinearRegression, train/test split)
[ ] Model Evaluation (R2, RMSE, MAE)
[ ] Deployment (Web UI/API endpoint)
[ ] Tools/Libs annotation (Pandas, scikit-learn, Jupyter)