MutagomaP/Vehicle_Analytics_Dashboard

Interactive Django dashboard for a synthetic vehicle dataset, featuring ML-powered price prediction, client segmentation, and folium maps (Rwanda districts + global country distribution). ## Vehicle ML Dashboard (Django + Scikit‑learn) This project is a small end‑to‑end machine learning dashboard built with **Django** and **scikit‑learn**. It loads a synthetic vehicle dataset, trains several ML models, and exposes them through a simple web UI with **EDA tables** and **geospatial visualizations**. ### Main Features - **Exploratory Data Analysis (EDA)** - Dataset head and basic description rendered as HTML tables. - Implemented in `predictor/data_exploration.py` and shown on the `/data_exploration/` page (`index.html`). - **Regression model – Price prediction** - Predicts vehicle selling price from features such as `year`, `kilometers_driven`, `seating_capacity` and `estimated_income`. - Trained in `model_generators/regression/train_regression.py` and served via `predictor/views.regression_analysis`. - Uses an `r2_score` (expressed as a percentage) to evaluate performance. - **Classification model – Price segment** - Classifies a vehicle into price/segment classes. - Trained in `model_generators/classification/train_classifier.py` and served via `predictor/views.classification_analysis`. - Uses `accuracy_score` (percentage) to evaluate performance. - **Clustering model – Client segmentation** - Unsupervised clustering of clients based on multiple numerical features such as `estimated_income`, `selling_price`, `kilometers_driven`, and `year`. - Implemented in `model_generators/clustering/train_cluster_improved.py`. - Uses **KMeans** on standardized numerical features and computes a **Silhouette Score** to quantify how well-separated the discovered clusters are in this feature space. - Clusters are mapped to human‑friendly classes: `Economy`, `Standard`, `Premium`. - Exposed through `predictor/views.clustering_analysis`. - **Geospatial visualizations** - **Rwanda map** (`predictor/rwanda_map.py`) - Uses a local GeoJSON file `predictor/data/rwa_adm2_simplified.geojson`. - `create_rwanda_map_with_districts(df)`: - Coun …