6 habilidades que todo Engenheiro de Dados deve ter re6eeornwy4ai9oeecmhyndhdktozn74kvmjguiob8

How a neutral fiber-optic network company in Brazil reached 93% accuracy in demand forecasting with a machine learning algorithm

See how beAnalytic’s data solutions supported the company’s strategic management

Industry

Technology

Sector

Fiber-optic distribution

Challenge

Demand forecasting

Solution

Machine Learning

The company

The client is a neutral fiber-optic network company, bringing high-quality connectivity to thousands of Brazilian households. It supplies infrastructure to several carriers and internet providers, democratizing access to cutting-edge technology in municipalities outside the major urban centers.

The challenge

The project’s main goal was to build a Machine Learning model able to predict when a CTO (a telecom distribution box, or transformer box) would reach its usage limit, that is, its maximum number of possible connections, forcing the company to install a new CTO in the region to meet demand. This way, the company’s team could work preventively and not only reactively when installing CTOs.

The data our client provided to train the ML model included internal information about the boxes as well as information from the local energy company, bringing in energy consumption data for the region.

The results

Accuracy in demand forecasting

We completed the creation and validation of the ML model within the 2-month period, reaching more than 93% accuracy in demand forecasting for our client.

Data pipeline

We created triggers between Cloud services, integration scripts and our production environment so the client could feed the database with new files on their own, without depending on beAnalytic.

Tracking dashboard

At the end of the process we built a dashboard in Power BI, which updates automatically to show the new forecast data in the best possible visual form.

 

Innovation case at OiWeek: we earned a top 5 spot in the 100 Open Startups ranking

We were invited to present this case atthe 2023 edition of 100 Open Startups’ OiWeek, where we could share a bit of how a Machine Learning project works on the inside.

This case also gave us another important milestone: we took 5th place in the Big Data category of the 100 Open Startups 2023 ranking.

See the event photos alongside →

foguete1

Find out how we can also help you cut costs and improve results

See other case studies of
companies transformed by beAnalytic

case 2 app

How a major delivery app cut its Cloud costs by 40% in Business Intelligence

case 3 logistica

How a last-mile logistics operator in Mato Grosso cut its losses by US$480k annually from damaged goods

case 4 ponto eletronico

How a time and attendance company eliminated 95 hours of monthly work by automating reports

Machine
Learning

Com a consultoria em Machine Learning da beAnalytic, a nossa equipe fica responsável por:

Mapeamento, coleta e tratamento dos dados necessários para o projeto;

Definição do algoritmo apropriado com base nos objetivos do projeto, e início do treinamento do algoritmo;

Avaliação do desempenho do modelo de ML, otimização e implementação no ambiente de produção.

A