Machine Learning Projects
Building intelligent systems through advanced machine learning techniques.
Movie Recommendation System
A content-based movie recommendation system using natural language processing techniques to suggest similar movies.
View Code
Sentiment Analysis
A natural language processing (NLP) model for analyzing sentiment in customer reviews.
View Code
Movie Recommendation System
• Data Preparation:
The process begins by loading and merging two CSV files—one with movie metadata and another with credits—using pandas. Unnecessary columns are dropped, and relevant features like cast, crew, keywords, and genres are retained to form a clean dataset.
• Building the Recommendation System:
Movie features (top 3 cast members, director, keywords, genres) are extracted, cleaned, and combined into a metadata "soup." This soup is transformed into numerical vectors using CountVectorizer for similarity computation.
• Getting Recommendations:
Cosine similarity is calculated between movie vectors, and a get_recommendations function uses these scores to return the top 10 movies similar to a given title, such as "The Dark Knight Rises."
Sentiment Analysis
• Data Preprocessing:
A dataset of over 14,000 tweets is loaded, neutral sentiments are filtered out for binary classification (positive/negative), and text is tokenized and padded to uniform length using TensorFlow's Tokenizer and padding utilities.
• Building the Text Classifier:
An LSTM-based neural network is constructed with an embedding layer, LSTM layers with dropout to prevent overfitting, and a dense layer with sigmoid activation for binary sentiment prediction.
• Training and Evaluation:
The model is trained for 5 epochs with a batch size of 32 and a 20% validation split, achieving 96% training accuracy and 94.33% validation accuracy, with a function provided to predict sentiment on new text inputs.