maithilipawar/travel-mode-predictor

Travel Mode Prediction Model

This model predicts the preferred travel mode (Walk, Bike, Car, Bus, Train, Flight) based on various factors including distance, budget, time constraints, and more.

Model Details

• Model Type: RandomForestClassifier
• Library: scikit-learn
• Input Features:
  • start_location: Starting location (categorical)
  • end_location: Destination location (categorical)
  • total_distance_km: Distance in kilometers (numerical)
  • season: Travel season (categorical)
  • day_type: Weekday/Weekend (categorical)
  • traffic_density: Traffic density (0-1, numerical)
  • user_budget: Budget in rupees (numerical)
  • user_time_constraint_hr: Time constraint in hours (numerical)

Usage

import joblib
import numpy as np
import pandas as pd

# Load model and pipeline
model = joblib.load('model.pkl')
pipeline = joblib.load('pipeline.pkl')

# Example input
input_data = {
    'start_location': 'Delhi',
    'end_location': 'Mumbai',
    'total_distance_km': 1400,
    'season': 'Summer',
    'day_type': 'Weekday',
    'traffic_density': 0.5,
    'user_budget': 5000,
    'user_time_constraint_hr': 24
}

# Preprocessing logic (simplified - see app.py for full implementation)
features = pipeline['features']
input_row = []

for feature in features:
    if feature == 'start_location':
        encoder = pipeline['label_encoders']['start_location']
        encoded = encoder.transform([input_data[feature]])[0] if input_data[feature] in encoder.classes_ else 0
        input_row.append(encoded)
    elif feature == 'end_location':
        encoder = pipeline['label_encoders']['end_location']
        encoded = encoder.transform([input_data[feature]])[0] if input_data[feature] in encoder.classes_ else 0
        input_row.append(encoded)
    elif feature == 'season':
        encoder = pipeline['label_encoders']['season']
        encoded = encoder.transform([input_data[feature]])[0] if input_data[feature] in encoder.classes_ else 0
        input_row.append(encoded)
    elif feature == 'day_type':
        encoder = pipeline['label_encoders']['day_type']
        encoded = encoder.transform([input_data[feature]])[0] if input_data[feature] in encoder.classes_ else 0
        input_row.append(encoded)
    else:
        input_row.append(input_data[feature])

# Scale numerical features
numerical_cols = ['total_distance_km', 'traffic_density', 'user_budget', 'user_time_constraint_hr']
numerical_data = {col: [input_row[features.index(col)]] for col in numerical_cols}
df_numerical = pd.DataFrame(numerical_data)
numerical_scaled = pipeline['scaler'].transform(df_numerical)

# Reconstruct full feature array
input_scaled = []
num_idx = 0
for feature in features:
    if feature in numerical_cols:
        input_scaled.append(numerical_scaled[0][num_idx])
        num_idx += 1
    else:
        input_scaled.append(input_row[features.index(feature)])

input_scaled = np.array([input_scaled])

# Predict
prediction = model.predict(input_scaled)
predicted_mode = pipeline['target_encoder'].inverse_transform(prediction)[0]

probabilities = model.predict_proba(input_scaled)[0]
mode_names = pipeline['target_encoder'].classes_
prob_dict = {mode: float(prob) for mode, prob in zip(mode_names, probabilities)}

print(f"Predicted mode: {predicted_mode}")
print(f"Probabilities: {prob_dict}")

Training Data

Trained on a dataset of travel scenarios across Indian cities with various travel parameters.

Files

• model.pkl: Trained RandomForestClassifier model
• pipeline.pkl: Preprocessing pipeline with encoders and scaler

Performance

The model achieves good accuracy on the test set for travel mode prediction based on the given features.