Skip to content
md ~/projects/sosyal-medya-reklam-analizi
Project archive

2026

AI-Powered Social Media Ad Analytics

  • Python
  • Scikit-learn
  • Pandas
  • NumPy
  • Streamlit
  • Plotly
  • SQLite
View on GitHub

Overview

Predicts whether a user will buy from a social media ad based on age and salary. Compares KNN, SVM, Logistic Regression and MLP models, tunes KNN with GridSearchCV and k-fold cross-validation, and serves results through a bilingual Streamlit interface.

01 · Problem

Finding out which classification algorithm works best on an ad dataset usually means trying each model in a separate notebook, which makes results hard to compare and hard to show to non-technical people.

02 · Solution

I built a Streamlit app that trains KNN, SVM (RBF kernel), Logistic Regression and MLP models on the same data and compares them. KNN hyperparameters are tuned with GridSearchCV and k-fold cross-validation; confusion matrices and metric charts are drawn with Plotly, and predictions are saved to SQLite.

03 · Outcome

Models can be compared side by side on accuracy, precision, recall and F1, and live predictions can be made from the interface. Built as a Machine Learning course project, with Turkish/English and light/dark theme support.