AI-Based Sentiment Analysis for Customer Reviews

Overview

The project aims to develop an AI-based sentiment analysis system using customer reviews scraped from Booking.com.

This repository is developed as part of the course M. Grum: Advanced AI-Based Application Systems at Potsdam University. It showcases an end-to-end AI project that involves data scraping, preparation, training, validation, and deployment using Docker and TensorFlow. The goal is to create an AI-driven system that is fully reproducible and can be utilized by researchers and evaluators.

---

Subgoals & Implementation

Subgoal 1: Git Usage

• Forked the MarcusGrum/AI-CPS repository and modified it to fit our project needs.
• Maintained structured commits with meaningful messages (at least three per team member).
• Documented project ownership in this README.md file and clarified that this project is part of the Advanced AI-Based Application Systems course.

Subgoal 2: Data Scraping & Preparation

• Scraped relevant data from the web and stored it in joint_data_collection.csv.
• Preprocessed data, including outlier removal and normalization.
• Split data into:
  • training_data.csv (80% of the dataset)
  • test_data.csv (20% of the dataset)
  • activation_data.csv (single entry from the test dataset)

Subgoal 3: Docker Images for Training & Activation

• Created two Docker images:
  • learningBase_SentimentAnalysis: Contains training data (training_data.csv) at /tmp/learningBase/train/ and test data (test_data.csv) at /tmp/learningBase/validation/.
  • activationBase_SentimentAnalysis: Contains activation data (activation_data.csv) at /tmp/activationBase/.
• Based images on BusyBox for lightweight deployment.
• Documented dataset origin and licensing in README.md inside Docker images.
• Verified functionality using docker-compose.yml and mounted external volume ai_system.

Subgoal 4: AI Model Training & Visualization

• Developed an AI model using TensorFlow.
• Trained the model on training_data.csv and validated it using test_data.csv.
• Stored the trained model as currentAiSolution.h5.
• Saved training metrics, loss, and accuracy plots.
• Created visualization reports including:
  • Training and testing loss curves.
  • Diagnostic plots.
  • Scatter plots.

Subgoal 5: OLS Model for Comparison

• Developed an Ordinary Least Squares (OLS) model using statsmodels.
• Trained and tested it using the same dataset for performance comparison.
• Stored the OLS model as currentOlsSolution.pkl.
• Documented performance using:
  • Diagnostic plots.
  • Scatter plots for analysis.

Subgoal 6: AI Model Deployment via Docker

• Created two additional Docker images:
  • knowledgeBase_SentimentAnalysis: Contains the AI/OLS models at /tmp/knowledgeBase/.
  • codeBase_SentimentAnalysis: Provides activation data for AI inference.
• Documented ownership, course affiliation, model type, and AGPL-3.0 license in README.md inside Docker images.
• Published images on Docker Hub for accessibility.

Subgoal 7: Docker-Compose Utilization

• Developed docker-compose.yml files for:
  1. Running the AI model using knowledgeBase_SentimentAnalysis and activationBase_SentimentAnalysis.
  2. Running the OLS model using the same setup.
• Used external volume ai_system for managing temporary files.
• Ensured seamless model execution by mounting required paths.

---

Setup & Usage

1. Clone Repository

git clone https://github.com/NoveraNasa/AI-Based-Sentiment-Analysis.git
cd AI-Based-Sentiment-Analysis

Course Information This repository is created and maintained by Qazi Novera Tansue Nasa and Mustafa Wasif as part of the course 'M. Grum: Advanced AI-based Application Systems' by the Junior Chair for Business Information Science, esp. AI-based Application Systems at University of Potsdam.

📄 License This project is licensed under the AGPL-3.0 license.