Create README.md

Includes detailed project description, model overview, input/output schema, and API usage notes. Part of Data Science Sprint 1 deliverables.

Author: ruvinyaye

Use_Cases/Environmental Impact Analysis/Ruvinya-Ekanayake/README.md

@@ -0,0 +1,80 @@
+Environmental Impact Analysis 
+
+Environmental Impact Analysis – CO₂ Savings Prediction
+This use case predicts how much CO₂ emissions can be saved when replacing an Internal Combustion Engine (ICE) vehicle with an Electric Vehicle (EV).
+The model compares EV energy consumption with ICE fuel consumption and estimates grams of CO₂ saved per km.
+A positive value means the EV emits less CO₂ than the ICE vehicle.
+
+1. Data Sources
+The model uses five datasets:
+|Dataset	                        |Description|
+|Pure electric consumption.csv	  |EV energy usage (Wh/km)|
+|Diesel consumption.csv	          | Diesel ICE fuel usage|
+|petrol91RON consumption.csv	    | Petrol 91 ICE consumption|
+|petrol95RON consumption.csv	    |Petrol 95 ICE consumption|
+|petrol98RON consumption.csv	    |Petrol 98 ICE consumption|
+
+2. Feature Engineering
+Key engineered features:
+|Feature	                |Meaning|
+|EV_gCO2_per_km	          |EV emissions converted from Wh/km → gCO₂/km|
+|ICE_CO2_Baseline	        |Estimated CO₂ for each ICE vehicle (fuel × emission factor)|
+|YearDiff	                |EV model year − ICE model year|
+|CO2_saving	              |ICE_CO2_Baseline − EV_gCO2_per_km|
+Emission factors used:
+•	Petrol: 23.2 kg CO₂ per L
+•	Diesel: 26.5 kg CO₂ per L
+•	Electricity: 0.18 kg CO₂ per kWh
+
+3. Modeling Approach
+Three models were tested using 5-fold cross-validation:
+|Model	            |Mean R²	       |Mean MAE|
+|Linear Regression	|Low	           |Moderate|
+|Random Forest	    |Moderate	       |Good|
+|Gradient Boosting	|Best	           |Lowest error|
+
+Selected model: GradientBoostingRegressor
+
+4. Production Model
+The production code:
+•	Loads and cleans data
+•	Combines EV + ICE using Cartesian merge
+•	Encodes categorical variables with OneHotEncoder
+•	Builds a pipeline with preprocessing + model
+•	Saves the trained model as:
+•	co2_savings_model.pkl
+
+5. Input Schema (API request)
+Example JSON input:
+{
+  "Make_EV": "Tesla",
+  "Make_ICE": "Toyota",
+  "BodyStyle_EV": "SUV",
+  "BodyStyle_ICE": "SUV",
+  "FuelType_ICE": "Petrol95",
+  "YearDiff": 5,
+  "ICE_CO2_Baseline": 220.4
+}
+
+6. Output Schema (API response)
+Example:
+{
+  "Predicted_CO2_Savings": 134.72
+}
+This means the EV is predicted to emit 134.72 g/km less CO₂ than the ICE vehicle.
+
+7. FastAPI Endpoint (Draft)
+Endpoint structure:
+POST /predict
+Accepts JSON input → returns predicted CO₂ savings.
+
+8. How the App Team Will Use This
+The mobile/web app will:
+1.	Send EV + ICE vehicle details to the API
+2.	Receive predicted CO₂ savings
+3.	Display environmental benefits to the user
+
+
+
+
+