Project Name

Predict daily total bike rentals and identify the key drivers of demand using historical daily bike-sharing data (day.csv). The analysis cleans the data, performs EDA, and builds Linear Regression models to produce actionable business recommendations.

Table of Contents

• General Info
• Technologies Used
• Conclusions
• Acknowledgements

General Information

• Background: Daily bike-sharing usage for two years was analyzed to forecast demand and surface operational drivers.
• Business problem: Forecast daily rentals (target cnt) and determine which factors (weather, calendar, environmental) most influence demand to support staffing, inventory, and pricing decisions.
• Observations: 731 daily records covering 2018–2019.

Conclusions

EDA summary

• Temperature is strongly positively correlated with bike rentals.
• Humidity and windspeed are negatively correlated with rentals.
• Clear/partly cloudy weather and Fall season show higher demand; heavy rain/snow drastically reduces demand.
• Demand is higher on working days and non-holidays; peak months: April–October.
• Noticeable increase in overall rentals in 2019 vs 2018.

Modeling summary

• Methods: simple OLS on single predictors and multiple OLS models with feature scaling and selection.
• Simple regression results (illustrative):
  • cnt vs temp: R² ≈ 0.59 (strong single predictor).
  • cnt vs humidity: R² ≈ 0.22.
  • cnt vs windspeed: R² ≈ 0.10.
• Multiple models implemented:
  1. Unscaled full feature OLS (baseline)
  2. StandardScaler + RFE (10 features)
  3. MinMaxScaler + RFE (10 features) — recommended
  4. MinMaxScaler + statistical p-value selection (≈16 features)
  5. StandardScaler + statistical selection

Best model & key features

• Recommended model: MinMaxScaler + RFE selecting 10 features.
  • Performance (reported in notebook): R² ≈ 0.82, Adj R² ≈ 0.82, RMSE ≈ 809, MAE ≈ 624.
• Top positive drivers: temp, yr=2019, certain season / month dummies (peak months).
• Top negative drivers: hum (humidity), windspeed, adverse weathersit (rain/snow), holidays / December.

Business recommendations

• Scale staffing and bike availability on warm, clear days (especially Apr–Oct and Fall).
• Monitor short-term forecasts for temperature, humidity, and windspeed to adjust daily operations.

Technologies Used

• Python 3.x
• pandas
• numpy
• matplotlib
• seaborn
• sklearn

Acknowledgements

• This project was based on Bike Sharing Rental Assignment.

Contact

Created by [@Divya3006] - feel free to contact me!