This repository contains the final project of Deep Learning course.
The course teaches students to build deep learning models for solving business problems using python libraries (e.g., Keras, TensorFlow, etc.). It covers a range of algorithms from neural networks foundations, to convolutional and recurrent network structures. The course will expose students to prevalent business applications of deep learning in different domains (Customer Analytics, Supply Chain Analytics, Healthcare Analytics, Financial Technology Analytics, Accounting Analytics, Talent Analytics, etc.). Upon completing this course, students will know how to build and optimize deep learning models for different business applications.
- Fundamentals of Deep Neural Networks; Applying Neural Network Models for Tabular Data Classification;
- Optimizing Deep Neural Networks; Applying Neural Network Models for Tabular Data Regression;
- Deep Learning for Computer Vision; Applying Convolutional Neural Networks for Computer Vision;
- Applying Sequential Models for Text Analysis; Fundamentals of Recurrent Neural Networks;
- Intro to Self-Supervised Learning and Generative Models; Applying Self-Supervised Learning for Clustering;
- Goal: Predicting the numerical rating based on the text-format review is our team project target.
- Data Source: The "Disneyland Reviews" dataset contains more than 40,000 reviews of the Disneyland Park in California, Tokyo and Hongkong. The reviews were scraped from TripAdvisor and cover a period of several years.
- Model Framework: First we did data preprocessing, normalizing and tokenizing. Then we set up 3 models, a LSTM model and a CNN model and compared them. Finally we adjusted parameters built the final rating prediction model.
- Model Accuracy: The mape as 4.89 and val_mape as 16.4140. The accuracy is 0.589 and precision is 0.58.
Course Instructor: Salih Tutun, PhD