Data Mining Assignment - Airline Passenger Satisfaction Prediction using Akkio Auto AI
People & Blogs
Introduction
In this article, I will discuss my recent assignment focused on predicting airline passenger satisfaction using Akkio, a no-code Auto AI platform. The process began with logging into the Akkio dashboard and uploading a dataset for analysis.
Dataset Overview
The dataset I used was downloaded from Kaggle, consisting of approximately 100,000 rows and 24 columns. Each row represents a specific passenger's travel journey, containing metadata such as the traveler's gender, customer type, and type of travel—be it personal or business. This comprehensive dataset provided a solid foundation for building an effective prediction model.
Data Cleaning and Preparation
The first critical step in developing our model was data cleaning. Akkio automated much of this process by standardizing the data columns and removing unexpected values. After cleaning the data, I converted categorical variables into numerical formats. For instance, I translated "female" to 0 and "male" to 1 for the gender column. Similarly, I converted "loyal customer" to 1 and "disloyal customer" to 0 for the customer type. This conversion to integer values was crucial for enhancing the model's prediction accuracy.
Model Training and Insights
In the prediction tab of Akkio, I used the satisfaction column as the target variable for classification. Other variables, including gender, age, and flight distance, served as predictors to ascertain their impact on passenger satisfaction. Of the 100,000 rows in the dataset, 80% were allocated for model training, while the remaining 20% were utilized to evaluate the model's accuracy post-training.
The insights obtained from the model were quite revealing. For example, the analysis showed that online boarding significantly influenced satisfaction results. Similarly, another impactful factor identified was the class of travel—business or economy.
Model Deployment
Following the successful model training, I proceeded to deploy the model. Akkio provides flexibility to select any number of columns to predict customer satisfaction levels. After removing "type of travel" and "class" columns, I updated the deployment settings.
Next, I uploaded a sample CSV file containing passenger data, excluding the satisfaction column, which the model would predict. Upon completion, the model generated satisfaction predictions for each entry in the uploaded file. For instance, a female passenger aged 52 was predicted to be neutral or dissatisfied with her experience.
Exploring Data Insights
Akkio also offers an exploratory feature to delve deeper into the dataset. I learned that the average satisfaction rating is approximately 1.43, assuming "neutral or dissatisfied" equals 1 and "satisfied" equals 2. Additionally, I discovered that flight distance does have an impact on passenger satisfaction, with the model providing further valuable insights.
In summary, this assignment has allowed me to better understand the data mining process, from data cleaning and model training to deployment and interpretation of results.
Keywords
- Airline Passenger Satisfaction
- Akkio Auto AI
- Data Cleaning
- Model Training
- Predictive Analytics
- Customer Insights
- Kaggle Dataset
- Data Exploration
FAQ
Q1: What is Akkio Auto AI?
Akkio is a no-code Auto AI platform that allows users to build predictive models without extensive programming knowledge.
Q2: How was the dataset used in the assignment obtained?
The dataset was downloaded from Kaggle and comprises travel journey data from approximately 100,000 passengers.
Q3: What steps were taken to prepare the data for model training?
Data cleaning was performed to standardize columns and remove erroneous values, followed by converting categorical variables into numerical values for better model compatibility.
Q4: What insights were gained from the predictive model?
The model revealed that factors such as online boarding and class of travel significantly impact passenger satisfaction rates.
Q5: How does the model predict satisfaction for new data?
The model is deployed to accept passenger data inputs and predicts satisfaction levels based on previously learned patterns.