Description

Descriptions

User reviews are a crucial source of information to understand customer satisfaction with Gojek services, such as GoRide, GoFood, and GoSend. Through these reviews, Gojek can identify its strengths and weaknesses. Sentiment analysis helps process these reviews into valuable insights to improve service quality.

Project Objectives

1. Analyze user review sentiment to identify positive, negative, and neutral trends.

2. Provide actionable insights to enhance the customer experience.

3. Create engaging data visualizations to simplify the understanding of analysis results.

Reference Studies

This project refers to several prior case studies, such as sentiment analysis of other app reviews on the Play Store or studies on ride-hailing services in Southeast Asia.

Project Workflow

Project Phases

1. Data Collection:

- User review dataset from Gojek is downloaded from Kaggle (containing reviews, ratings, and service categories).

- Dataset Features:

• Content : User reviews in text format.
• Score : Star rating from 1 to 5.
• At: Review date.
• Username : Gojek user account name

2. Data Preprocessing:

- Data Cleaning:

• Remove unnecessary symbols, numbers, and punctuation.
• Convert text to lowercase

- Tokenization:

• Split sentences into individual words.

- Stopword Removal:

• Remove common words like "yang," "di," and "dan" that do not add significant information.

- Stemming:

• Reduce words to their root form (e.g., "makanan" to "makan").

- Sentiment Encoding/Labeling : 

• Categories based on ratings:

1–2: Negative.

3: Neutral.

4–5: Positive.

3. Exploratory Data Analysis (EDA):

- Analyze sentiment distribution based on ratings

- Visualize frequently occurring words using WordCloud

- Analyze positive and negative word frequency.

4. Modeling:

• Vectorization Technique: TF-IDF is used to convert text into numerical representations.
• Machine Learning Models: Algorithms used: Random Forest and Support Vector Machine (SVM)

5. Model Evaluation:

- Evaluation Methods: Accuracy, Precision, Recall, and F1-Score.

- Compare model performance and select the best model.

Insights & Conclusion 

Key Insights

1. Top Positive Words: "easy," "fast," "good" indicate that users appreciate the efficiency of the service.
2. Top Negative Words: "slow," "expensive," "error" highlight key issues that need improvement.
3. Sentiment Distribution: The majority of positive reviews show customer satisfaction, but there are significant complaints about delivery time.

Recommendations for Gojek

1. Improve Response Time: Focus on reducing delivery time and fixing app bugs.
2. Optimize Pricing: Evaluate pricing policies to be more competitive.
3. Real-Time Dashboard: Develop a real-time analytics system to monitor customer sentiment directly.