Sentiment Analysis of Food Estate Merauke

Description

This notebook contains an implementation of sentiment analysis on the topic of Food Estate Merauke using text processing techniques and machine learning classification. The analysis aims to understand public opinion regarding the Food Estate policy developed in Merauke, Indonesia.

The processes conducted in this notebook include:

1. Data Processing:

   • Loading a dataset of comments related to Food Estate obtained through YouTube scraping.
   • Shuffling the dataset to avoid bias.

2. Data Preprocessing:

   • Text Cleaning: Removing numbers, symbols, brackets, and emoticons.
   • Word Normalization: Converting informal words or abbreviations into their correct form.
   • Stopword Removal: Eliminating common words that do not have significant meaning in sentiment analysis.
   • Stemming: Reducing words to their root form using Sastrawi.

3. Sentiment Analysis:

   • Implementing a classification model to determine positive, negative, or neutral sentiment.
   • Evaluating model performance using specific metrics (accuracy, precision, recall).

---

Sentiment Analysis Results

Confusion Matrix and Model Evaluation

The Confusion Matrix illustrates the model's performance in classifying sentiment into three categories: Negative (-1), Neutral (0), and Positive (1).
Based on the model evaluation, the following metrics were obtained:

• Model accuracy of 80%, indicating the model's correctness in predicting sentiment.
• F1-score for each class:
  • Negative (-1): 0.70
  • Neutral (0): 0.76
  • Positive (1): 0.86
    The model performs best in identifying positive sentiment, while negative sentiment has a lower recall.

Sentiment Distribution

The pie chart displays the sentiment distribution of classified comments:

• 60% of comments are positive, indicating that most users support or agree with the Food Estate program in Merauke.
• 22.2% of comments are neutral, meaning they do not express a clear opinion on the policy.
• 17.8% of comments are negative, suggesting dissatisfaction or criticism from some members of the public.

Conclusion

• Most YouTube comments show positive sentiment towards Food Estate Merauke.
• The classification model used has a fairly good accuracy (80%) in classifying sentiment.
• The presence of negative sentiment indicates that some concerns or criticisms about the program still exist.

---

Suggestions for Improvement:

• Further exploration of different models could help compare and optimize performance.
• Using a larger dataset could improve model generalization.
• Adding visualizations of sentiment analysis results could provide clearer insights into public opinion trends on Food Estate Merauke.