Preprocessed the netflix movie data with NTLK module. Next, I trained a variety of Deep Learning model for clustering movies and classifying them based on generated label. Used LSTM and BERT for model training.
Got a F1-Score of 0.97 using KMeans model for clustering and Pre-trained BERT and BiLSTM model for classification.
Proposed a Bi-LSTM sequential model consisting of multiple dense layers with numerous nodes, using a ReLU activation function. L2 regularization was used to handle class imbalance.
{Evaluated combinations of hyper-parameters (primarily used Learning Rate, Nodes per BiLSTM layer, Number of BiLSTM layers, and Dropout after BiLSTM layers) to acquire the best model with an F1-Score of 0.84.
Trained various models, namely CNN+SVM, CNN+LSTM+SVM, and pre-trained BERT-based models on Kaggle’s News Headlines Dataset.
Acquired the maximum accuracy of 98\% with the pre-trained BERT model and an F1-Score of 0.96.