Analyzing the Sentiment Behind British Airways Customer Assignment

Assignment Task

Introduction

1.1 Background

According to Shakya, Du and Ntalianis (2023), Sentiment analysis is an analysis that includes sentiment extraction in the process of determining the attitudes and feelings related to the textual or informational input. Nowadays, people depend on reviews and personal recommendations. aviation industry's competitive market has grown rapidly during the last couple of decades. This analysis can predict how the public will react to the company and can determine whether or not the clients are satisfied with the given pricing and service. Therefore, in this case, a large number of good reviews have a huge impact on both passengers and the airline company. Therefore, people will naturally opt for items with a more significant number of positive reviews. Similar to this, products with terrible opinions will try to discourage customers from using them, so that it will cost money to the enterprise (Xu, Zhang and Lin 2023).

1.2 Problem statement

According to Lei and Liu (2021), in the dynamic landscape of the airline industry, know-how and addressing purchaser sentiment is paramount to keeping aggressive and building brand loyalty. However, the traditional way of manually combing through copious amounts of purchaser comments and reviews is not only time-consuming but also prone to subjectivity. This poses a massive challenge for airlines seeking to derive actionable insights from the wealth of data at their disposal. For instance, the rapid proliferation of online platforms and social media has exponentially increased the volume and variety of patron comments, making sentiment reading a nightmare. Therefore, there arises an urgent need for more effective and objective techniques to sentiment analysis within the airline quarter. Although language representation models present a bright prospect towards this task, implementing these within the scope of airline patron comments presents its own set of challenges (Zapata, Javier and Carlos 2018).

1.3 Research questions

1) How effectively can language representation models analyze sentiment in consumer comments and reviews for airline services?

2) What are the most frequently experienced emotions stated by customers in airline comments and reviews?

3) How do sentiment trends vary among airline services, including booking, in-flight amenities, and customer service?

4) What factors influence customer feedback and reviews on airline services?

1.4 Research aim & Objectives

The aim of this study is to use language representation models to do sentiment analysis on customer feedback and reviews for airline services, revealing trends, patterns, and factors that influence sentiment.

Objectives

1) Evaluate language representation models for sentiment analysis of airline service comments.

2) Identify common attitudes in consumer comments and reviews.

3) Analyze sentiment patterns for airline services.

4) Identify elements that impact consumer attitude in airline service feedback.

5) Make recommendations for airlines to improve customer happiness based on sentiment analysis results.

1.5 Conceptual framework

In the context of sentiment analysis for airline customers' feedback, the dependent variables would therefore include sentiments left by the passengers, demographic data on the passengers like age, gender, nationality, frequent flyer status, and source of income; mentioned specific service attributes within the feedback like in-flight amenities, comfortability of the seat, staff behavior, and cleanliness. The variables in this case are therefore indicators of passenger satisfaction and preferences that impact on their overall experience and probability of repeat business.

1.6 Limitation of the research

According to Wankhade, Rao and Kulkarni (2022), There are certain disadvantages to utilizing language illustration models for sentiment analysis in the aviation industry. First, these fashions are subject to biases discovered in the training data, which may result in prejudiced or incorrect sentiment forecasts. Furthermore, it is fundamentally difficult to convert model results into a clear form that can help determine the underlying causes of high sentiment returns in customer comments. Furthermore, language representation techniques may struggle to capture the whole range of linguistic intricacies and cultural nuances seen in airline customer evaluations, particularly those written in languages other than English. Given the givens on aspects of privacy and consent, one should be very careful when collecting and reading consumer comments. Facts, safety tips, and ensured privacy of individuals should be kept in mind for accountable and ethical research into this field (Holthoff 2016).

02. Methodology

2.1 Dataset

For this study, airline reviews were collected from Kaggle. Information submitted by passengers or travelers who have shared their thoughts or emotions about the airlines they have been on is displayed. Categories include "good," "negative," or "neutral" feelings about the airline, a list of airlines that passengers have flown on, traveler and passenger information, location, time zone, and the text or remark from the passenger are all included in the data.

2.2 Estimators Used in Data Analytics

In this report, we discuss the estimators used for sentiment analysis on airline reviews. The primary estimators considered are Neural Network, Logistic Regression, and K-Nearest Neighbors (KNN). Each estimator offers unique strengths and limitations, making them suitable for different aspects of the analysis.

1. Neural Network (NN):

Neural networks are powerful models inspired by the human brain's structure. They consist of interconnected nodes organized into layers, with each node applying a mathematical operation to its inputs and passing the result to the next layer.

Strengths

• Neural networks can capture complex relationships in data, making them suitable for sentiment analysis, where sentiments may be nuanced and context-dependent.
• They can automatically learn features from the data, reducing the need for manual feature engineering.
• Neural networks can handle high-dimensional data effectively.

Limitations

• Training neural networks requires a large amount of data and computational resources.
• They are prone to overfitting, especially when dealing with small datasets or complex architectures.
• Tuning hyperparameters can be challenging and time-consuming (Alzahrani and Alenazi 2021).

2. Logistic Regression (LR):

Logistic Regression is a linear model used for binary classification tasks. It models the probability that a given input belongs to a particular class using a logistic function.

Strengths

• Logistic Regression is computationally efficient and interpretable, making it suitable for quick analysis and model interpretation.
• It performs well when the relationship between features and target variable is linear or close to linear.
• Logistic Regression can handle both numerical and categorical features.

Limitations

• Logistic Regression assumes a linear relationship between features and target variable, which may not hold true for complex data.
• It may underperform when dealing with highly nonlinear relationships.
• Logistic Regression is sensitive to outliers and multicollinearity among features (Liu and Lang 2019).

3. K-Nearest Neighbors (KNN):

K-Nearest Neighbors is a non-parametric algorithm used for classification tasks. It classifies a new data point based on the majority class of its K nearest neighbors in the feature space.

Strengths

• KNN is simple to understand and implement, making it suitable for quick analysis or as a baseline model.
• It makes no assumptions about the underlying data distribution, allowing it to capture complex patterns.
• KNN can handle multi-class classification tasks and works well with small to medium-sized datasets.

Data analysis and discussion

The initial performance metrics for the binary classification problem, analyzing the sentiment of British Airways passenger reviews using Neural Network, Logistic Regression, and K-Nearest Neighbors, revealed slight differences in performance at the first epoch. We observed that Neural Network posted a better training loss compared to Logistic Regression and KNN, which were slightly above. However, all the models posted very low validation losses in the first epoch, meaning that they all learned efficiently. In addition, Logistic Regression posted the lowest validation loss, followed closely by KNN and then Neural Network.

A comparative analysis showed parallel performances between Neural Network, Logistic Regression, and KNN in the binary task, with slight variations in accuracy and other metrical scores. However, Logistic Regression demonstrated slightly better precision scores for predicting 'Negative' sentiment compared to Neural Network and KNN (Wankhade, Rao and Kulkarni 2022).

Our results have shown major improvement in accuracy for Neural Network, Logistic Regression, and KNN, where all have surpassed the 95?curacy benchmark. Logistic Regression performed best among all the models tested for the binary classification task, scoring the highest accuracy. Furthermore, its best performance is accentuated by leading in all three metric measurements across both classes, testifying to its effectiveness in performing sentiment analysis tasks.

We found that the accuracy has drastically improved in the case of Neural Network, Logistic Regression, and K-Nearest Neighbors. Each model surpassed the 95?curacy benchmark set. In this binary classification task, Logistic Regression appeared to perform better than any of the models considered, with approximately 96.97?curacy. The prowess of Logistic Regression in the metrics measure is further highlighted by its top score in all three metrics across both classes, peaking at approximately 98.78% for the 'Negative' label.

This Accounting and Finance has been solved by our PHD Experts at My Uni Paper.