End-to-end machine learning project for bike demand prediction using EDA, feature engineering, regression modeling, and validation. Includes business insights and data-driven strategies for optimizing fleet allocation, pricing, and operations based on weather, seasonality, and user behavior

This project analyzes a bike-sharing dataset to understand demand patterns and build a predictive model using Linear Regression. The goal is to identify key factors influencing bike rentals and derive actionable business insights.

• Perform Exploratory Data Analysis (EDA)
• Identify key factors affecting bike demand
• Build a regression model to predict bike rentals (cnt)
• Validate model assumptions
• Generate business insights & strategies

The dataset contains daily bike rental data with features such as:

• 🌡️ Temperature (temp, atemp)
• 💧 Humidity (hum)
• 💨 Windspeed (windspeed)
• 🌦️ Weather situation (weathersit)
• 📅 Date, month, weekday
• 👥 Casual & registered users
• 🎯 Target: Total rentals (cnt)

• Distribution plots for numerical features
• Boxplots for categorical variables vs demand
• Correlation heatmap
• Pairplot for feature relationships

• Temperature strongly influences demand
• Bad weather reduces usage significantly
• Demand varies across seasons and months

• Converted categorical variables (month, weekday, weather)
• Created dummy variables
• Removed multicollinearity (temp, casual, registered)
• Feature scaling using MinMaxScaler
• Train-test split (70:30)

• Recursive Feature Elimination (RFE)
• Statistical significance (p-values)
• Variance Inflation Factor (VIF)

• yr, atemp, windspeed, season_spring, mnth_Jul, weathersit_C

• Linear Regression (Statsmodels & Sklearn)
• Multiple models compared:
  • 15 features
  • 7 features
  • 6 features (final optimized)

• Residual analysis → approximately normal
• Residual vs predicted → slight heteroscedasticity
• VIF → low multicollinearity
• Strong generalization on test data

• Distribution plots
• Correlation heatmap
• Residual plots
• Actual vs Predicted scatter plot

• Perceived temperature (atemp)
• Year-over-year growth

• Humidity
• Windspeed
• Bad weather conditions

• High demand: Fall, moderate weather
• Low demand: Spring, extreme summer

• Use weather-based prediction models

• Increase prices in high demand
• Discounts in bad weather

• Allocate bikes based on demand patterns

• Boost demand in low seasons

• Schedule maintenance in low-demand periods

• Python
• Pandas, NumPy
• Matplotlib, Seaborn
• Scikit-learn
• Statsmodels

├── data.csv

├── Smart_Mobility_Prediction.ipynb

├── Data dictinonary.txt

├── README.md

git clone https://github.com/your-username/bike-demand-analysis.git
cd bike-demand-analysis
pip install -r requirements.txt
jupyter notebook