Big Data Based Points of Interest Analysis

HIGHLIGHTS

Duration: 5 months
Industry: Marketing and Advertising
Location: USA

ABOUT CLIENT

The client is an innovative advertising company delivering cutting edge technology solutions for marketing and advertising in malls and supermarkets across the US.


BUSINESS CONTEXT

The customer created an alerting system for sellers which was based on the developed algorithm, and it functioned to inform that a potential client is approaching the store location. Aterise was to create an algorithm that detects users’ movement around the mall and recognize typical routs and points of interest.

CALLENGE

The core task for Aterise team was to develop a neural network algorithm that would analase the collected Big Data and transform it to hunman-readable actionable insights to increase the quality of customer service at hypermarkets and malls.

From the user application standpoint, at this stage Aterise team was required to create a fully functioning MVP relying on the provided mockup by the customer. The prototype of the system contained 3 potential user roles: the administrator, the user, and the business user. Besides the opportunity to upload the examples of AI algorithms to the website, it was necessary to implement the data verification system of the algorithm.

SOLUTION

Aterise developed a functional frontend system with an opportunity for further backend integration based on the customer’s mockup. As a result, we delivered software that predicts people’s movements along the area with beacons placed around. The software uses data taken from the accelerometer and a user phone’s compass with the software set up on the phone.

  • Aterise not only turned the customer’s idea into reality but also managed to cut down their expenses for front-end and back-end integration
  • We managed to train the algorithm on the basis of the data received from the user’s device (cell phone). That resulted in the ability to predict user movements with an accuracy of 3 meters.
  • We optimized the data, trained the neural network algorithm, and achieved prediction accuracy of up to 97%.
  • Finally, Aterise conducted a test check of the solution beginning from the simplest neural network algorithms up to the most complex algorithms. Thus, we made the system operate accurately by creating an ideal algorithm.


Thank you for reading!

EXPLORE OTHER STORIES →

GET IN TOUCH

You've landed in the right place, let us know what can we do for you!

    We will process your personal information in accordance with our Privacy Policy.