Predicting Fiber Uptake in an Area Using Machine Learning

Local Chapter Eldoret, Kenya Chapter

Coordinated byKenya ,

Status: Completed

Project Duration: 26 Jul 2023 - 11 Aug 2023

Open Source resources available from this project

Project background.

The project’s background stems from the growing importance of reliable and high-speed internet connectivity in Kenya, particularly with the increasing number of individuals and businesses relying on digital technologies and remote jobs.

According to a report from the Communications Authority of Kenya, home fiber usage by Kenyans surpassed mobile data subscriptions, which experienced a 0.7% decline in Q3 of 2019/2020. Home data and fiber-to-the-office constitute the largest share of broadband users.

Understanding the fiber uptake rate in different areas is crucial for telecommunication companies and policymakers to effectively plan and allocate resources for expanding fiber networks.

The problem.

Telco companies face a lack of data-driven and quantitative tools to assist in their fiber roll-out strategies, particularly in predicting household fiber uptake rates. Existing solutions primarily concentrate on cost calculations from providers like Comsof and Netadmin, rather than providing comprehensive support in identifying potential expansion areas.

Developing a tool that offers data-driven revenue predictions based on specific areas can significantly enhance the accuracy of expansion business cases. Gaining insights into the projected fiber uptake in households can serve as a valuable initial reference for decision-making.

Project goals.

- A model that can be able to predict the uptake rate of fiber. - Visual map showing the most attractive fiber roll-out areas.

Project plan.

  • Week 1

    Data Collection – Identify and gather relevant datasets. i.e Demographic, Competitive fiber, and satellite datasets.

  • Week 2

    1. Process Data – Clean and aggregate data into one granular unit.
    2. Exploratory Data Analysis

  • Week 3

    1. Set up structure of model 2. Train and test prediction model based on demographical and test data

  • Week 4

    Design and create interactive user interface (map and dashboard)

Learning outcomes.

1. Data Collection – Web Scraping
2. Data Modeling
3. Machine learning
4. Data Visualization in a map.
5. Geospatial Data.
6. Research

Share project on: