Sentiment Analysis of Fintech Application Reviews

To develop a web-based sentiment analysis application integrated with data mining using the CRISP-DM framework. This project supports my undergraduate thesis and enables users to automatically analyze the latest user review sentiments and discover frequently discussed topics through topic modeling.

Client: Thesis

Service: Machine Learning Engineer

Date: 26 January 2025

Users: Open to the public, mainly used for academic research purposes.

Provides an automated and efficient way to classify user review sentiments and extract key topics without manual processing.

Developed an automated web-based sentiment analysis system using Python and Streamlit

Fetched up to 2,000 latest app reviews dynamically via API from the UI

Performed automated text preprocessing, dynamic train-test splitting, and labeling using pretrained BERT

Trained sentiment classification using SVM and predicted test data with word cloud visualization

Automatically generated topic modeling from positive sentiment results

Secure login system with authentication and role-based access control

Web-based application integrated with machine learning.

Streamlit

Python

BERT (Bidirectional Encoder Representations from Transformers)

Algorithm Support Vector Machine (SVM)

Latent Dirichlet Allocation (LDA)