Developed a complete machine learning pipeline to predict engagement on social media posts using structured data. This project demonstrates data cleaning, feature preparation, preprocessing pipelines, model training, and evaluation in scikit-learn.

Key Contributions:

• Dataset: Social media posts dataset with features such as platform, post type, post time, likes, comments, shares, and sentiment.
• Exploratory Data Analysis (EDA): Inspected distributions, correlations, and engagement patterns by platform and post type.
• Data Preprocessing:
  • Scaled numeric features (likes, comments, shares) with StandardScaler.
  • Encoded categorical features (platform, post_type, post_day, sentiment_score) using OneHotEncoder.
• Pipeline Construction: Built a full scikit-learn Pipeline combining preprocessing with model training.
• Model Training:
  • Tested multiple models, including Logistic Regression and Decision Tree Classifier.
  • Evaluated with accuracy scores on test data.
• Reflection: Explained design choices for feature prep, imputation (none required), and model selection.

Skills Demonstrated:

• Machine Learning (classification pipelines)
• Feature scaling, encoding, and preprocessing with ColumnTransformer
• Pipeline design & evaluation in scikit-learn
• EDA and visualization with Pandas, Matplotlib, Seaborn
• Structured documentation of methodology & reasoning

Links:

• GitHub Repository 
• Rendered ipynb Notebook PDF