TripRobotics Develops a Price Predictor Tool Capable of Forecasting Price Movements

Background

Our customer is an OTA which previously replaced an older version of the booking service with TripRobotics’s help.

Yet another project needed an upgrade and it was a Price Predictor. To accomplish the task, TripRobotics had to implement a fare forecast on top of the existing search engine. The product is meant to help customers make better purchasing decisions by displaying the best possible time to make the purchase over several months.

Objectives

Having extensive experience in the travel market, our client decided to step further and implement the already existing in the industry Price Predictor Tool to provide its users with increased customer service and better price saving opportunities. Our mission included the creation of an intelligent algorithm, able to forecast price movement based on data and smart analysis methods.

By accomplishing the mission, TripRobotics has established the following objectives:

  • In-depth analysis of data and broad research
  • Development of the prediction algorithm
  • Optimization of the user experience and performance

Deliverables

Revealing the Hidden Patterns in Broad Dataset
To be able to understand and visualize the hidden patterns, TripRobotics team had to overtake several processes that included historical data analysis, handling of neighbor travel dates data and merging the algorithms to form the time series for further forecasting. The information acquired by analyzing millions of searches allowed our team to make highly accurate predictions. There were also employed advanced data mining and aggregation techniques.

Developing the Future Price Movements Feature
Following discovering the patterns, our team was able to form the algorithm models with different parameters. As we got the info about actual flight fares, we were able to select the convenient algorithm: based on the past data, our data analysts were capable of running the algorithm and observing the correlation between the actual data and the obtained predictions. The achieved confidence rate constituted 75%, based on analysis for long-term (7 weeks) and short-term (7 days) predictions. The algorithm is improved using machine learning techniques and factual information on correct and wrong predictions.

Smooth Price Predictor Integration and Optimization
The Price Predictor is currently integrated and is displayed for about 20% of users in a separate popup window and a search module to test the interaction between the newly introduced elements and users. This allows our team to capture the necessary stats and improve the features based on the received reactions and information, thus to improve the customer experience. For instance, the average time per session has doubled once the price predictor was integrated.

Defined Delivery Process
By keeping an ongoing integration process and testing the features meticulously, the design flaws and defects were considerably minimized. As a consequence, the delivery process became efficient and predictable. A modern version control system (Git) has been integrated which led to simple and convenient code management.

Technical Info and Methods

The total duration of the project: 6 months.

Professionals Involved: Data scientist, UX/UI designer, and software engineers.

Technologies Included: R programming language, C#, Data Mining, Data Aggregation and Extrapolation, Time Series Forecasting.

Wondering how to build
an exclusive software product for your business?