AI-Projects

Projects to implement various Machine Learning algorithms.

1. Perceptron Classifier for Hotel Reviews: Implemented a perceptron-based classifier (Vanilla Perceptron and Averaged Perceptron) for classifying hotel reviews(taken from TripAdvisor, Expedia).

   • Vanilla Perceptron: Achieved a mean F1 score of Pos F1: 0.92 Neg F1: 0.92(Sentiment of reviews) and True F1: 0.86 Fake F1: 0.86 (Validity of Hotel reviews). Mean F1 score: 0.8874

   • Averaged Perceptron: Achieved a mean F1 score of Pos F1: 0.92 Neg F1: 0.92(Sentiment of reviews) and True F1: 0.87 Fake F1: 0.88 (Validity of Hotel reviews). Mean F1 score: 0.8968

   Programming Language: Python

2. Naive Bayes Classifier for Hotel Reviews: Implemented a Naive Bayes classifier to classify Hotel reviews (taken from TripAdvisor, Expedia) based on two categories:

   • The validity of the hotel reviews: True/Fake

   • The sentiment of the hotel reviews: Pos/Neg

   Achieved a mean F1 score of Pos F1=0.94, Neg F1=0.94, True F1=0.88, Fake F1=0.89, Mean F1 = 0.9124

   Programming Language: Python

3. Collegiate Football Predictions: A Neural-Net model to predict football winning bets using tensor flow and Numpy.

   • Utils functions: To determine the best network architecture and parameters to get an optimal MSE.

   Programming Language: Python

4. First Order Logic Inference Engine: Agent to implement first-order logic inference engine using resolution and unification algorithms.

   • Parsing of the logical statement in CNF form and creating a knowledge base.

   • Applying unification algorithm to clauses in the knowledge base to resolve the query to determine entailment.

   Programming Language: Python

5. Minimax Game Playing Agent: A Minimax game playing agent which uses alpha-beta pruning to search for varying depths of moves to find an optimal next move for the user.

   Programming Language: Python

6. Built a Part-of-Speech Tagger for machine translation of English(89.94% accuracy), Hindi(92.41% accuracy), and Chinese (86.55% accuracy) languages.

   • Developed a Hidden Markov Model for estimating the transition and emission probabilities.

   • Viterbi Decoding to generate the most likely sequence of tags for word sentences.

   Programming Language: Python

7. Animal Placement in a Nursery: Extension of N – Queens: Implementation of global search strategies like Depth First Search, Breadth First Search and Simulated Annealing to place K animals in a N * N nursery with obstructions.

   Programming Language: CPP