This repository contains a semantic analysis of a Net Promoter Score (NPS) survey, focusing on survey scores, reasons for the scores, and suggestions for improvements.
The analysis involves data preprocessing, topic modeling using Gibbs Sampling Dirichlet Mixture Model (GSDMM), coherence scoring, and visualization.
- Python Libraries: Pandas, NLTK, Gensim, Matplotlib
- Data: Analysis conducted on
Score_Alldata.csv.
The code for this project is divided into several key parts, each responsible for a stage in the analysis process:
-
Data Preprocessing: Cleaning and tokenising text data. This step is crucial as it prepares the raw survey text, removing any noise and irrelevant information. Tokenisation is particularly important because it breaks down the text into individual elements or 'tokens'. This makes the data more manageable and enables the model to more effectively analyse and understand the language patterns and sentiments in the responses.
-
Topic Modeling: Applying GSDMM for identifying topics within survey responses. We chose GSDMM specifically because it excels in analysing short texts less than 50 words, which is typical in this survey setting. This model helps in accurately extracting themes and sentiments from concise responses, providing a deeper understanding of the underlying topics in the survey data.
-
Visualisation: Generating word clouds and bar charts to visualise the findings. These visualisations are essential for presenting the results in a way that is easily understandable, even for those with a non-technical background. For example, the word cloud below shows the most frequently mentioned terms in one of the survey categories, offering a quick and intuitive visual representation of key themes.
Each of these steps plays a vital role in transforming raw survey data into meaningful insights, enabling us to understand customer opinions and preferences better.
Based on our analysis, we've uncovered key insights into what drives customer satisfaction and areas for improvement:
- Customers frequently mention signal quality and network services, especially in lower scores, indicating a major area of concern.
- Positive notes in higher scores revolve around good service and value for money, suggesting these are strong points.
- Improvement suggestions are mainly focused on enhancing signal coverage, service quality, and internet speed.
- Improve Network Services: Prioritise upgrading signal strength and network coverage.
- Maintain High-Quality Customer Service: Continue investing in customer service to keep satisfaction levels high.
- Communicate Proactively: Inform customers about improvements being made in response to their feedback.
- Deepen Analysis: Enhance the analysis of survey feedback using advanced techniques.
- Data Integration: Merge NPS data with other customer information like purchase history and service interactions for comprehensive insights.
- Text Mining and Sentiment Analysis: Apply text mining on open-ended responses and use sentiment analysis to uncover underlying emotions and viewpoints.
- Predictive Analytics: Utilise machine learning to predict future trends and behaviors based on current NPS data, identifying potential areas of focus.
- Customer Engagement: Reach out to respondents for more in-depth feedback, especially those who gave lower scores.