📊 Sentiment Analysis on GLO Telecommunication Feedback

This project conducts a comprehensive sentiment analysis on public complaints and customer feedback collected from GLO Nigeria subscribers. Using machine learning and NLP techniques, the project classifies user comments into Positive, Neutral, or Negative sentiments, across four key service categories:

• Customer Service
• Data Plan
• Quality of Service (QoS)
• Tariff Plan

---

🧠 Project Objective

To gain deep insight into the perception of GLO services by analyzing textual customer comments, and generate data-driven recommendations for enhancing customer satisfaction.

---

📁 Project Structure

Sentiment-Analysis/
│
├── Sentiment_Analysis_Glo.ipynb             # Main notebook with analysis and visualizations
├── Sentiment_Machine_Learning_Approach.ipynb# Alternate ML model implementations
├── Sentiment_Model.ipynb                    # Model tuning and versioning
│
├── Glo_customer_services_comment.csv        # Customer service comments
├── Glo_data_plan_comment.csv                # Data plan feedback
├── Glo_qos_comment.csv                      # Comments on quality of service
├── Glo_tariff_plan_comment.csv              # Tariff plan complaints
│
├── README.md                                # Documentation

---

📊 Sentiment Distribution Results (by Category)

🧑‍💼 Customer Service

• Negative: 152 (70.3%)
• Positive: 48 (22.2%)
• Neutral: 16 (7.4%)

📌 Majority of feedback highlights poor responsiveness, rude agents, and unresolved complaints.

---

📶 Data Plan

• Negative: 159 (72.9%)
• Positive: 45 (20.6%)
• Neutral: 14 (6.4%)

📌 Key concerns include fast data depletion, expensive bundles, and poor value for money.

---

📡 Quality of Service (QoS)

• Negative: 160 (72.7%)
• Positive: 43 (19.5%)
• Neutral: 17 (7.7%)

📌 Users frequently complain about slow internet, frequent call drops, and network instability.

---

💰 Tariff Plan

• Negative: 166 (76.1%)
• Positive: 38 (17.4%)
• Neutral: 14 (6.4%)

📌 Users expressed dissatisfaction with unfavorable call rates, billing errors, and lack of transparency.

---

🔧 Tools & Technologies Used

• Python
• Pandas, NumPy – data cleaning and structuring
• Scikit-learn – model training (Logistic Regression, Naive Bayes)
• NLTK – text processing
• TF-IDF – feature vectorization
• Matplotlib / Seaborn – data visualization
• WordCloud – visual sentiment keywords

---

🧪 Workflow Summary

1. Data Preprocessing

• Cleaned and normalized text
• Tokenization, stopword removal, lemmatization
• Labeled datasets (positive, neutral, negative)

2. Vectorization

• TF-IDF applied to transform text into features

3. Model Building

• Logistic Regression & Naive Bayes used for classification
• Model tested on all 4 categories independently

4. Evaluation

• Accuracy, Precision, Recall, F1-score
• Confusion matrix and sentiment distribution visualized

---

📈 Insights & Recommendations

• Sentiments are heavily skewed toward negative across all categories.

• GLO’s customer experience strategy needs revamping, especially around:

  • Agent support training
  • Transparent billing
  • Better internet service coverage

🎯 Targeted service improvements can reduce churn and improve brand perception.

---

🚀 How to Use

1. Clone the repository:

   git clone https://github.com/Percy-O/Sentiment-Analysis.git
   cd Sentiment-Analysis

2. Install dependencies:

   pip install -r requirements.txt

3. Launch the main notebook:

   jupyter notebook Sentiment_Analysis_Glo.ipynb

4. Step through the cells to view preprocessing, analysis, and final results.

---

📦 Requirements

You can create a requirements.txt like:

pandas
numpy
scikit-learn
matplotlib
seaborn
wordcloud
nltk

---

📬 Contact

Author: Percy Owoeye GitHub: @Percy-O Email: owoeyepercyolawale@gmail.com