Trip-duration-prediction-using-NN

This is a simple project to predict the approximate time during a taxi trip.

The input includes the latitude & longitude of the pick-up & drop-off sites and the pick-up time.

The output predicted time is in the form of '0-5min', '5-10min', '10-15min', '15-20min', '20-15min', '25-30min, '>30min'

How to run

To run this project, you can follow the steps below.

To run predictions using the pre-trained model:

The model parameters have be saved in model1_10epoch.h5 file.

I've set the input to be the original test data, so just python run.py .

Training the model from scratch

• Download the New York City Taxi Trip Duration Dataset and store it in a directory dataset. Here I just used the train.zip as the whole training and testing dataset. Please change the datapath(the path for train.csv) in utils.py according to your path.
• python train.py
• python run.py

Data Preprocessing

These following features are extracted as inputs of the neural network

• Distance features: 'distance_haversine', 'distance_dummy_manhattan'
• Speed features：'avg_speed_h'，'avg_speed_m', unit:m/s
• Time features: 'pick_up_h', 'pick_up_m', 'weekday', (0 represents Sunday)
• Zone features: 'pickup_lat_label', 'pickup_long_label', 'dropoff_lat_label', 'dropoff_long_label'

Neural Network

A very simple Fully Connected Network with 3 Dense layers and 2 Dropout layers and SGD optimizer