Previsao de Churn de Clientes

Customer Churn Prediction

Quais clientes estao prestes a sair, e da pra saber antes? Construi um modelo que aprende os sinais de quem esta de saida, do tratamento dos dados ate a avaliacao. Vendo isso com olhos de quem ja cuidou de carteira: prever churn e proteger receita.

Which customers are about to leave, and can we know before they do? I built a model that learns the signals of someone on their way out, from data cleaning to evaluation. I see this through the eyes of someone who managed accounts: predicting churn is protecting revenue.

Player Churn Prediction

Player Churn Prediction

Como gamer, quis entender por que jogadores abandonam um jogo, e juntei o util ao que amo. Um modelo de classificacao que prediz a evasao de players, aplicando analytics a dados de gaming.

As a gamer, I wanted to understand why players leave a game, so I combined the useful with what I love. A classification model that predicts player churn, applying analytics to gaming data.

Modelagem de Dados & ETL

Data Modeling & ETL

Antes de qualquer grafico bonito, alguem precisa arrumar a casa. Montei um pipeline de ETL que transforma dados brutos e baguncados em estruturas confiaveis, a base silenciosa de todo relatorio que se sustenta.

Before any pretty chart, someone has to tidy up the house. I built an ETL pipeline that turns raw, messy data into reliable structures, the silent foundation of every report that holds up.

Modelo de Regressao — Previsao de Vendas

Regression Model — Sales Forecasting

Quanto vamos vender no proximo mes? Passei anos respondendo isso no feeling; agora respondo com estatistica. Um modelo de regressao que projeta vendas a partir do historico, ligando numero a decisao comercial.

How much will we sell next month? I spent years answering that on gut feeling; now I answer it with statistics. A regression model that forecasts sales from history, linking the number to a commercial decision.

Commerce Intelligence Report

Peguei ~80 mil transacoes de e-commerce e construi o painel que eu gostaria de ter tido na epoca de vendas: visao executiva, canais e perfil de cliente, com KPIs e filtros. Dados de estudo, mas a leitura e de quem viveu a mesa comercial. Construido em React.

I took ~80k e-commerce transactions and built the dashboard I wish I'd had back in my sales days: executive overview, channels and customer profile, with KPIs and filters. Study data, but the reading comes from someone who lived the commercial desk. Built in React.

Segmentacao de Clientes (RFV)

Customer Segmentation (RFV)

Nem todo cliente e igual, e eu sei disso na pratica. Usei RFV e K-Means para separar a base em quatro perfis, dos campeoes aos que estao em risco, e dei a cada grupo um nome de negocio e uma acao. Clusterizacao com cabeca comercial.

Not every customer is the same, and I know that from practice. I used RFV and K-Means to split the base into four profiles, from champions to at-risk, and gave each group a business name and an action. Clustering with a commercial mind.

Airbnb — Limpeza & Analise

Airbnb — Cleaning & EDA

Dados do mundo real chegam baguncados, e tudo bem. Limpei um conjunto do Airbnb tratando valores ausentes e outliers, e fui atras das historias escondidas nos numeros: o que move preco e ocupacao.

Real-world data arrives messy, and that's fine. I cleaned an Airbnb dataset, handling missing values and outliers, then went after the stories hidden in the numbers: what drives price and occupancy.

Analise de Dados Salariais

Salary Data Analysis

O que de fato pesa no quanto alguem ganha? Analisei dados salariais para encontrar os fatores por tras da remuneracao, com visualizacoes e estatistica que traduzem o numero em conversa.

What really weighs on how much someone earns? I analyzed salary data to find the factors behind compensation, with visualizations and statistics that turn the number into a conversation.

Customer Churn — Pipeline ML

Customer Churn — ML Pipeline

Aqui fui mais fundo no churn: pipeline completo com engenharia de atributos, comparacao de varios modelos e avaliacao por metricas, sempre perguntando o porque, nao so o quem. Entender o motivo da evasao vale mais que so prever.

Here I went deeper into churn: a full pipeline with feature engineering, comparing several models and metric-based evaluation, always asking the why, not just the who. Understanding the reason for churn is worth more than just predicting it.