Navigating Airline Customer Feedback:
Sentiment Analysis and Topic Modeling with Gensim LDA and K-means
The airline industry is highly competitive and customer service is a critical area that can influence customer loyalty and brand reputation. Timely and effective handling of customer feedback is crucial for maintaining service quality. Analyzing customer feedback can provide insights into common concerns and areas for improvement.
Project Goals
1. Sentiment Analysis: To automatically determine the sentiment of customer comments. The sentiment can be:
- Negative (Complaint): Indicates the customer faced an issue or is dissatisfied.
- Positive (Congratulation): Reflects customer satisfaction or a positive experience.
- Neutral (Opinion): Represents a customer's opinion that is neither a direct complaint nor praise.
2. Complaint Classification: To classify customer feedback into categories for routing to the appropriate department. This classification will help in streamlining the process of addressing customer feedback efficiently. Example categories include:
- Luggage Issues: For complaints or feedback related to damaged or lost luggage.
- Flight Delays: For issues pertaining to delayed or cancelled flights.
- Billing Issues: For concerns related to overcharges, refunds, or payment discrepancies.
The end product will be a model or a set of models that can accurately tag incoming customer comments with the correct sentiment and category, allowing the customer service team to promptly address and resolve the issues.
Sentiment Analysis
I conduct sentiment analysis on customer feedback comments received by an airline to understand the public perception and identify areas for improvement. The goal is to categorize each comment into positive, neutral, or negative sentiment, providing actionable insights that can help the airline enhance its service quality and customer satisfaction. For this purpose, I have utilized the TextBlob library, a popular Python tool for processing textual data. TextBlob simplifies text processing by offering a straightforward API to conduct sentiment analysis, among other NLP tasks. It internally uses the Natural Language Toolkit (NLTK) for Python and leverages pre-trained models to assess the polarity of sentiments expressed in the comments. By applying TextBlob, I can efficiently process large volumes of customer feedback, enabling the airline to swiftly identify trends in customer sentiment and respond to their needs more effectively
I focused on visualizing the distribution of customer contact channels for the airline company. This chart illustrates the proportion of comments received through each channel, including Email, Facebook, Outbound Email, Phone, Portal, and Twitter.
I generated a bar chart to depict the frequency of comments received on each day of the week. It presents a straightforward comparison of daily comment counts, offering insights into the peak and trough periods of customer activity.
I created a heatmap to understand the temporal distribution of customer interactions with an airline company throughout a typical day and across different days of the week. It illustrates the intensity of customer activity per hour for each day.
Gensim LDA Model
After building an LDA (Latent Dirichlet Allocation) model with the Gensim library, the topics displayed the top keywords for each of the 5 topics identified by the model. This method showcases the most significant words that characterize each topic, helping to understand the thematic focus of these topics.
1. First item
2. Second item
- Subitem 2.1
- Subitem 2.2
3. Third item
K-means Model
After building an LDA (Latent Dirichlet Allocation) model with the Gensim library, the topics displayed the top keywords for each of the 5 topics identified by the model. This method showcases the most significant words that characterize each topic, helping to understand the thematic focus of these topics.
1. First item
2. Second item
- Subitem 2.1
- Subitem 2.2
3. Third item