Ola Bike Rides Request Demand Forecast

Ola Bike Rides Request Demand Forecast

Given big data at taxi service (ride-hailing) i.e. OLA, you will learn multi-step time series forecasting and clustering with Mini-Batch K-means Algorithm on geospatial data to predict future ride requests for a particular region at a given time.
explanation image

Videos

Each project comes with 2-5 hours of micro-videos explaining the solution.

ipython image

Code & Dataset

Get access to 102+ solved projects with iPython notebooks and datasets.

project experience

Project Experience

Add project experience to your Linkedin/Github profiles.

Customer Love

What will you learn

What is multi-step time series forecasting

Different strategies to solve Multi-Step Time Series problem explained
Converting a Time Series problem to Supervised Learning problem
Data Preprocessing in Time Series
Understanding the business problem - based on which process and prepare data
Feature Engineering in Time Series - breaking time to minutes, hours, day of the week, date, month
Geospatial Engineering - lat/long clustering
MiniBatchKmeans algorithm clustering on a large scale
Concept of Auto-Correlation Function (ACF) and Partial Auto-Correlation Function (PACF) in time series
Lead-Lag Concept
Rolling Mean Concept
Random Forest Algorithm
Xgboost Algorithm with lag features and rolling window mean
Implementation of recursive multi-step time series forecasting method
Evaluation Metrics: Root Mean Square Error

Project Description

Project Description

The taxi service (ride-hailing) industry is growing for the last couple of years and it is expected to grow in near future. Taxi drivers need to choose where to hang tight for passengers as they can get somebody quickly. Passengers also prefer a quick taxi service whenever needed. We all have faced problems with taxi booking requests, that sometimes cannot be fulfilled or the wait time for ride arrival is very long due to the unavailability of a nearby taxi. One should feel fortunate in the event that you get a taxi booked in one go.


Taxi demand prediction has become extremely important for taxi-hailing (and e-haling) companies to understand their demand and optimize their fleet management.

To handle such issues, we would be building a model based on users ride request dataset, which would contain attributes: ride booking time, pickup point, and drop point latitude-longitude. This model would forecast the demand, for a particular time in different areas of the city which would help the company optimize taxi concentration to fulfill users demand.

 

Business Problem

Ola Bikes are suffering losses and losing out from their competition due to their inability to fulfil the ride requests of many users. To tackle this problem, you are asked to predict demand for rides in a certain region and a given future time window. This would help them allocate drivers more intelligently to meet the ride requests from users.

 

Goal

You have to predict ride requests (demand forecast) for a particular latitude

and longitude for a requested future time window/duration.

 

Data Description

Raw Data contains a `number` (unique for every user), ride request DateTime (IST time),

pickup and drop location latitude, and longitude.

 

Data Fields

  1. number: unique id for every user
  2. ts: DateTime of booking ride (IST time)
  3. pick_lat: ride request pickup latitude
  4. pick_lng: ride request pickup longitude
  5. drop_lat: ride request drop latitude
  6. drop_lng: ride request drop longitude

 

Defining a Good Ride Request

Ola Management knows the task is not easy and very important for their business to grow.

Hence, their business team has provided you some guidelines to follow.

  1. Count only 1 ride request by a user, if there are multiple bookings from the same latitude and longitude within 1hour of the last booking time.
  2. If there are ride requests within 8mins of the last booking time consider only 1 ride request from a user (latitude and longitude may or may not be the same).
  3. If the geodesic distance from pickup and drop point is less than 50meters consider that ride request as a fraud ride request.
  4. Consider all ride requests where pick up or drop location is outside India bounding box: ["6.2325274", "35.6745457", "68.1113787", "97.395561"] as system error.
  5. Karnataka is our prime city where we have a lot of drivers and ride requests to fulfil. We would not love to serve rides that are outside Karnataka and have pickup and drop geodesic distance > 500kms. Karnataka bounding box: ["11.5945587", "18.4767308","74.0543908", "78.588083"]

 

Predict Task to test your model

After model development, Ola has requested us to build a prediction pipeline for the deployment of the model. To test our prediction pipeline, they have provided us clean data (filtered rides requests data based on good ride definition conditions) of 2021-03-26 based on which they have requested us to predict/show initial hours rides request demand forecast for 2021-03-27.

 

 

New Projects

Curriculum For This Mini Project

Understanding the OLA bikes demand prediction problem
07m
Project flow and structure
02m
Files and Folder Structure
05m
Multi Step time series forecasting intuition
05m
Data exploration and Basic Data cleaning
05m
Data analysis and Advanced data cleaning Part 1
05m
Data analysis and Advanced data cleaning Part 2
03m
Data analysis and Advanced data cleaning Part 3
04m
Data Preparation and Geospatial Engineering
05m
Data Preparation and Time feature engineering
02m
Model Building Part 1
05m
Lead and Lag in Time series
02m
Model Building Part 2
04m
Model Building Part 3
03m
Model Prediction Pipeline
05m

Latest Blogs