Original text

case studies

GeneralPublished on: Fri Nov 04 2022

Implementation of an Independent Telecom Analytics Platform for a Telecommunication Firm

Business Scenario

The client is a leading telecom provider firm based in the United States. They wanted to build an AI-enabled predictive analytics platform to generate reports for telecom KPIs and predictive analytics like the RSRP & RSRQ prediction, Signal Strength prediction, and Gap Fill prediction for Virtual Drive Tests of networks for any telecom market of the world.

Business Challenges/ Pain Points

The major challenge faced by our client was generating and analyzing the insights from various telecom providers. They had the data and insights only for the networks maintained by them & for the specific markets. The client wanted a platform to analyze the historical data from multiple sources and gain insights into customer behavior which can further help them provide a more customized experience through better prediction & forecasting.

Hexaview Solution

To meet the client’s requirement, Hexaview developed a Big Data Analytics and ML Predictive platform. The platform gathers raw data from multiple telecom crowd-data provider companies. This data is further parsed and stored in a data warehouse with the help of various big data processing tools like Hadoop, Hive, Impala, and Apache Spark.

The KPI prediction models were trained and deployed for Signal Strength Prediction and Gap Fill Prediction for Virtual Drive test networks.

Technology Stack

1. Languages Used

Backend:- Python, HDFS, Hive, Impala, Apache Spark, Tensorflow, cuDF, cuML

Frontend:- ReactJS, High Chart

2. Algorithm(s) Used – DBSCAN , SVM , RNN

3. Libraries – Apache Spark, Tensorflow, cuDF, cuML, Parquet

Amount of Data Handled in this Project 

Around 250- 300 GB of Crowd Data of the telecom industry is processed and maintained in the warehouse every day.

Impact on Business

  • Enabled the client to understand the customer behavior based on insights from data analytics.
  • The clients can now measure the engagement and identify the preferences of a particular user.
  • The client can now make predictions about the user behavior for the desired network KPI.