FARE PREDICTION SYSTEM

Yashvardhan Mahecha
Date: 12-08-2021
Abstract:
In travelling cost of a vehicle is a determinant of the time it takes to reach a certain destination
along with the factor that whether the route is busy or empty which itself is a determinant of the
environmental factors.

Predictive analytics is the use of data, statistical algorithms and machine learning techniques to identify
the likelihood of future outcomes based on historical data. The goal is to go beyond knowing what has
happened to providing a best assessment of what will happen in the future.

Though predictive analytics has been around for decades, it's a technology whose time has come. More
and more organizations are turning to predictive analytics to increase their bottom line and competitive
advantage. Why now?

1. Growing volumes and types of data, and more interest in using data to produce valuable
insights.
2. Faster, cheaper computers.
3. Easier-to-use software.
4. Tougher economic conditions and a need for competitive differentiation.

In this report we have devised an optimal way to predict the price of fare so as to get a glance about the
prices and manage finance accordingly.

1. Problem Statement
Providing a methodological approach to analyze the fare market in depth to get insights
about the determining factors for the increase in fare of a ride along with using the concept of
machine learning to train the model accordingly and predict whenever required.

2. Market/Customer Need Assessment

There has been a huge surge in the automobile sector and online food delivery system. Our aim
is to analyze the price prediction algorithm for the system which will help Individuals and MNC’s
to get a glance about fare predictions which will help them to plan the price accordingly.
2. With the growth of combining the concept of ‘Big Data and Machine-Learning’ and
introducing machine learning model for data analysis and the importance of data produced by
the cabs on daily basis and how this data can be used by the machine learning to tell the
consumer about the exact price of their ride before starting the ride. This provides the
consumer to make better choice of cab based on the price predicted by the Machine-Learning
model. The proposed system uses the cab dataset and weather dataset to make predictions for
each ride booked by the customer.
3. Target Specification and characterization
Finding and alter map routes based on the trends that have been proposed by the Machine
Learning algorithm via checking the price for the route.
If there is a subset of routes possible for a particular location, then our algorithm can be used to
find an optimal solution based on the predictions of prices which is made by our machine
learning model. If the value is lowest, that path will be chosen.
Price prediction will help individuals plan their journey according and can use the robust
model to analyze how far they can travel .

Getting the insights about price for travelling can help predicting how far an individual can
travel within a certain budget .

4. External Search(information sources)

The dataset can be found on the kaggle .The dataset consists about the distance between the
initial and final destination along with type of cab.The sources of subsequent information is
given below as referencs.

The weather dataset consists of the following entities:

5. Benchmarking
The above data gives us a glance about the different types of cab companies along with their
difference in numbers in the market along with the different source destination that the people
chose in their journey as well as the weather factors that may contribute to increased prices in
the use case
6. Applicable Patents
International Journal of Interdisciplinary Innovative Research &Development (IJIIRD)
ISSN: 2456-236X Vol. 05 Issue 01 | 2020 -The current patent may incorporate this patent for
the inspiration of the methodology used as well as EDA analysis to some extent.

7. Applicable Regulations
The patents mentioned above might claim the technology used if the algorithms are not
developed and optimised individually and for our requirements. Using a pre-existing model is off
the table if it incurs a patent claim.
1. Must provide access to the 3rd party websites to audit and monitor the authenticity and
behavior of the service.
2. Enabling open-source, academic and research community to audit the Algorithms and
research on the efficacy of the product.
3. Laws controlling data collection : Some websites might have a policy against collecting
customer data in form of reviews and ratings.
4. Must be responsible with the scraped data : It is quintessential to protect the privacy and
intention with which the data was extracted.

8. Applicable Constraints:
1. The use of cloud platforms to store the data gathered over the net.
2. Using the spark service to clean and transform data .
3. For Evaluation of the model which is done with the help of tableau and PowerBI.
4. For modelling using Timeseries(SARIMAX) and linear regression is applied.

9. Business Opportunity
9.1 Substantial amount of tourism opportunities can be seen if the stats can be used by
local tourism firm and government support (At least in India) presents us with a promising future
in product development and comparison. If Someone doesn’t have a lot of capital to spend on
market research and feedback on prototype products. Our service offers just that. We provide a
one stop price comparison, feedback, and comparative analysis all using Machine learning and
automation. This will complement their market research team gain better insights on product and
pricing design .
9.2 Transportation sector can utilize this segmentation report to charge for a particular
customer service that they desire.
10. Concept Generation
This product requires the tool of machine learning models to be written from scratch in
order to suit our needs. . Tweaking these models for our use is less daunting than coding it up
from scratch. A well trained model can either be repurposed or built. But building a model with
the resources and data we have is dilatory but possible. The customer might want to spend the
least amount of time giving input data. . This accuracy will take a little effort to nail, because it’s
imprudent to rely purely on Classic Machine Learning algorithm .

The Accuracy of the initial model is given below:

11. Concept Development

The concept can be developed by using The appropriate API (flask in this case) and using Django as
framework for the same and for its deployment , The cloud services has to be choosen accordingly to
the need.
12. Final Report Prototype
The product takes the following functions to perfect and provide a good result.
Back-end
Model Development: This must be done before releasing the service. A lot of manual supervised
machine learning must be performed to optimize the automated tasks.
1. Performing EDA to realize the dependent and independent features.
2. Algorithm training and optimization must be done to minimize overfitting of the model and
hyperparameter tuning.

Front End
1. Different user interface: The user must be given many options to choose form in terms of
parameters. This can only be optimized after a lot of testing and analysis all the edge
cases.
2. Interactive visualization the data extracted from the trained models will return raw and
inscrutable data. This must be present in an aesthetic and an “easy to read” style.
3. Feedback system: A valuable feedback system must be developed to understand the
customer’s needs that have not been met. This will help us train the models constantly.

13. Product details - How does it work?

An interactive user system will take inputs regarding the journey from the user and the user will get to
know about the price for the journey that he wanted in real time considering the weather factor in mind
regarding the same with the user interactive UI.

14. References/Source of Information

https://medium.com/@Vishwacorp/timeseries-forecasting-uber-demand-in-nyc-54dcfcdfd1f9

https://www.kaggle.com/ravi72munde/uber-lyft-cab-prices

https://blog.dataiku.com/predicting-taxi-fares-in-new-york-using-machine-learning-in-real-time

http://ijiird.com/wp-content/uploads/050144.pdf