https://github.com/dejongyeong/sentiment-analysis
Final Year Project: Sentiment Analysis Approach for Reputation Evaluation
https://github.com/dejongyeong/sentiment-analysis
hybrid-approach lexicon-based machine-learning naive-bayes-classifier python scikit-learn sentiment-analysis support-vector-machines
Last synced: 2 months ago
JSON representation
Final Year Project: Sentiment Analysis Approach for Reputation Evaluation
- Host: GitHub
- URL: https://github.com/dejongyeong/sentiment-analysis
- Owner: dejongyeong
- Created: 2019-01-30T22:15:21.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2020-10-25T13:44:26.000Z (over 4 years ago)
- Last Synced: 2025-04-12T05:00:00.225Z (2 months ago)
- Topics: hybrid-approach, lexicon-based, machine-learning, naive-bayes-classifier, python, scikit-learn, sentiment-analysis, support-vector-machines
- Language: Python
- Homepage:
- Size: 132 MB
- Stars: 5
- Watchers: 0
- Forks: 4
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Sentiment Analysis Approach for Reputation Evaluation
Link to slide: [SlideServe](https://www.slideserve.com/dejongy/sentiment-analysis-approach-for-brand-reputation-evaluation-powerpoint-ppt-presentation)
### Research Question:
* Can computers perform reputation evaluation that benefit businesses?
* How sentiment analysis can be used to evaluate reputation of a product or services?
### Objective
* Evaluate and Compare the selected sentiment classification techniques used to evaluate brand reviews.
* Findings are presented to make informative decisions regarding the adoption of classification techniques.
### Proposed Methodology
* Dataset containing 400,000 reviews of unlocked mobile phones sold on Amazon was selected.
* Three approaches (lexicon-based, machine learning and hybrid) were implemented to identify the underlying sentiment.
* Model evaluation metrics were utilised for comparative analysis.
### Programming Language and Software Tools
* PyCharm by JetBrains
* Python
* Scikit-Learn Library
### Results
* Accuracy of Hybrid approach was the highest, giving 81.2% of correctly predicted observation.
* Precision score of Lexicon-based approach was the lowest with 54.0% of correctly predicted positive observations.
* F1 score of Hybrid approach was the highest, presenting with 70.2% of harmonic mean between precision and recall.
* The positive sentiment label in Apple and BlackBerry mobile reviews were higher, compared to the negative and neutral sentiment labels.
### Conclusions
* Hybrid approach to sentiment analysis can effectively be used to evaluate brand reviews that benefit businesses.
* Underlying sentiment of brand reviews can be evaluated with the use of sentiment classification techniques.
* Slang and emoticons handling may be implemented to improve results of sentiment analysis.