With increasing connectivity, the quality of user experience (QoE) and efficient management of telecommunications network resources are crucial issues for telecom operators. In this context, Orange has undertaken a strategic project to optimize the dimensioning of its 4G/5G RAN access network using Big Data and Artificial Intelligence (AI). This initiative is part of the operator's strategic ambitions to offer exceptional connectivity and an unparalleled customer experience.
Project context and objectives
Our experts were called upon to evolve the methodologies for evaluating and predicting QoE and capacity suffering, i.e., congestion of the 4G/5G access network.
The main objectives of this project were:
- To ensure the management and evolution of the ARD Datalab.
- To develop a criterion for qualifying and analyzing customer behavior in terms of traffic flow and QoE customer satisfaction.
- To implement a detection grid using Big Data to identify congested sites according to the ICON criterion.
- To implement AI-based dimensioning methodologies for the 4G/5G access network for budget loops (4G/5G traffic forecasts, QoE, and capacity suffering).
- To provide multi-project strategic capacity and QoE studies for central and regional departments.
- To evaluate internal 4G access network dimensioning tools.
- To supervise the migration to more advanced solutions.
Methodology and organisation
The methodology adopted by Sofrecom is based on the exploitation of massive data (Big Data) and the application of AI models for network dimensioning. Here are the key steps of this methodology:
1. Data Extraction, transformation, and loading (ETL)
- Extraction of data from probes, topology, and customer information
- Transformation and loading of data into the ARD Datalab using Linux and Hive scripts
2. QoE KPIs creation
- Development of QoE KPIs per hour, day, month, and per cell, sector, site
- Collection, processing, and transformation of OSS data at Busy Hour (BH)
3. Forecasting and Modeling
- Use of AI models (Machine Learning and Deep Learning) to predict capacity needs and QoE
- Forecasting of OSS KPIs taking into account bandwidth additions, events, and school holiday periods
4.Visualization and Analysis
- Development of interactive maps and dashboards using Power BI
- Classification of ICON capacity suffering and extraction of relevant files
Tools and skills used
The tools and programming languages used in this mission include Hue, Edge Datalab (Linux), Power BI, Python, Pyspark, and Hadoop. The skills required to carry out this mission include data mining, Big Data analysis, Business Intelligence (BI), as well as Machine Learning (ML) and Deep Learning (DL).
Deliverables and results
The main deliverables of this mission include:
- A monthly ICON detection grid to identify congested sites
- Studies on seasonality and events
- An estimation of capacity needs
- Analyses related to 4G/5G services and offers
- Studies related to Green to reduce the carbon footprint without altering QoE.
The results obtained have made it possible to guarantee excellent QoS and QoE, optimize investments with a gain of 6 to 8%, and reduce the carbon footprint of the operator.
The dimensioning of the 4G/5G RAN access network using Big Data and AI is a strategic mission that has enabled the client to guarantee optimal connectivity and an exceptional customer experience. By exploiting massive data and applying AI models, Sofrecom has succeeded in providing accurate forecasts, optimizing investments, and reducing the carbon footprint, while ensuring premium QoE. This mission illustrates the importance of innovation and data analysis in the telecommunications sector to meet the growing needs of users and promote financial inclusion.
For more information on this mission and the solutions offered by Sofrecom, please do not hesitate to contact us, or visit our website.
