Skip to content
Snippets Groups Projects
Select Git revision
  • main default protected
1 result
Name Last commit Last update
Venv
actions
data
models
tests
Add_Fake_data.py
Airline_Dataset.csv
Chatbot_image.gif
FakeDataset.csv
Project_Video.mp4
README.md
airport_encoder.pkl
app.py
config.yml
country_code_encoder.pkl
credentials.yml
domain.yml
endpoints.yml
model_1.pkl
model_2.pkl
requirments.txt
train_models.py
README.md

Patel, Dhruvang, 22306142

Talaviya, Jevin, 22402981

Title - Airline enquiry

https://mygit.th-deg.de/dp08142/assistance-system

https://mygit.th-deg.de/dp08142/assistance-system/-/wikis/home

Project Description

This project involves analyzing Airline dataset containing records of passenger information, flight details, and airport data. The main objectives of this analysis are to explore trends, identify patterns, and derive insights that can enhance the understanding of airline operations and customer details.

Prerequisites

This project is develop using the following:-

  1. Python - 3.8.2
  2. Python virtual environment - 20.4.7
  3. System - Windows 11
  4. Streamlit Version - 1.40.1
  5. Pandas Version - 2.0.3
  6. Scikit- learn Version - 1.1.3
  7. Matplotlib Version - 3.5.3
  8. Rasa Version - 3.6.20
  9. Joblib Version - 1.2.0
  10. Altair Version - 5.4.1
  11. Faker - 33.3.1

Installation

  1. Clone the repository.
  2. Create a Virtual environment:> python -m venv Venv
  3. Activate the Virtual Environment: Venv\scripts\activate (Windows) or source/Venv/bin/activate (Linux / MacOS).
  4. Install Dependencies: pip install -r requirements.txt.
  5. Now Follow the steps of "Basic usage" to start the application.

Basic usage

  1. After setting up the Installation, Open the Terminal from Menubar or press 'CTRL + J'.
  2. Change Powershell to Comand Prompt by pressing the plus(+) symbol on top right corner of the Terminal.
  3. Activate the Virtual Environment (Venv) in Comand Prompt: Venv\scripts\activate (Windows) or source/Venv/bin/activate (Linux / MacOS).
  4. Now you are in your Virtual environment, Next step is to start Rasa server by: rasa run.
  5. Now it will show in console that Rasa server is up and running, It indicates that till now your rasa is running perfectly without any errors (You are very Lucky).
  6. If rasa server is running, Then Again open a new comand prompt from Terminal without closing the previous prompt.
  7. Now it's time to Run the chatbot on Streamlit. For that enter : streamlit run app.py in new Comand Prompt.
  8. This will direct you to a default Browser, where you will find our 'Airline Chatbot'.

Overview of Chatbot

  • Title and Branding - At the top, there is a welcoming gif with hello logo and the title "Airline Chatbot & Data explorer with predictions" indicating the chatbot's purpose.

  • Central Panel - When you scroll down, there is a section labeled "Chat with Airline Assistant". This allows user to interact with chatbot by entering specific details like Passenger ID & Last Name. Then on clicking on Get Flight Details it retrieves the passenger's flight information. Where it shows wheather the flight is On Time or Delayed and further information.

  • Features on the Left Sidebar - The left Sidebar has features like Chatbot, Data Explorer, Visualization and Predictions. Below there is Explanation on each features.

    1. Chatbot: - Likely the current mode, allowing users to interact with the chatbot for flight details or other queries.
    2. Data Explorer: - It explore the Airline Dataset. First one is Select columns to view where you can select on your own which columns information you want and second one is filter of flight status.
    3. Visualizations: - It show the visualizations of top 10 passenger nationalities, passenger age distribution and flight status distribution.
    4. Predictions: - It show the predictions of passenger satisfaction and flight status.

This layout appears user-friendly and efficient for retrieving passenger detials and exploring related data through multiple features.

Data

Data Source - We have use Airline Dataset from Kaggle.

Data Handling -

  1. Outlier Detection - The method that can handle both numerical and categorical data effictibely is Isolation Forest algorithm.
  2. Fake Data - Added 25% realistic data balancing across target variable.

There is more details explain about data in Wiki.

Implementation of the Requests

Here is the overview of the requests:

  • Web App with Streamlit: This project was made with the help of streamlit where it facilitate components like inputing messages, slidebar, data analysis, buttons etc.

  • requirments.txt: In this file it contains different version for dependencies. The file is available on git repository.

  • CSV file: The dataset was take from kaggle where you can find here Airline Dataset. It show the original version of dataset but later it was edit according to project use.

  • Pandas: With the help of pandas we loaded the csv file and implemented functions to locate the data of the users as well as for data overview and visualizations.

  • Scikt-Learn Model: The code uses RandomForestClassifier to train both models.

  • Matplotlib: With the help of matplotlib we have implemented graphs and pie charts according to out dataset.

  • Chatbot Implementation with RASA:

    1. Intent: domain.yml & nlu.yml file where it gives the intent and according to that it show the responses for the user & nlu gives the examples according to the intent.
    2. Actions: actions.py file contains the back-end data where it read the csv file and use class function to get passenger flight details from the csv file and generate the response to users.
    3. Streamlit File: app.py file contains the front-end data, which communicate between the user and the chatbot.
    4. stories.yml: stories.yml where it shows the process of conversation for the chatbot.
    5. rules.yml: rules.yml where user says goodbye and conversation end.
    6. endpoints.yml: endpoints.yml where things runs smoothly with the help of url.
    7. train models.py: train_models.py where it contains two models for the prediciton of the data.

Each component has it's own where it has created user- friendly chatbot and gives predictions and analysis according to the data.

Work Done

The Chatbot was done by two team members, each contributing their work in this project. Below it's show which member has done which work:

Student 1: [Patel Dhruvang, 22306142]

  1. Data analysis with pandas & Strategies for outliers and fake data

  2. Scikit-learn

  3. Dialog flow

Student 2: [Talaviya Jevin, 22402981]

  1. Graphical User Interface

  2. Visualization (with pandas and matplotlib)

  3. Sample Dialog

Both: Documentation, Rasa Implementation and Programming

Demonstrating Video

A video of the project is upload on git repository. If the video is buffering, please download the video and know how our Chatbot works.

Project Structure

 Airline Inquiry Chatbot/
 │
 ├── data/
 │   ├── nlu.yml/              
 │   ├── rules.yml/                
 │   ├── stories.yml/              
 │
 ├── actions/                       
 │   ├── _pycache_            
 │   ├── __init__.py       
 │   ├── action.py   
 │                               
 ├── Add_Fake_data.py
 │
 ├── Airline_Dataset.csv
 │
 ├── Chatbot_image.gif 
 │
 ├── FakeDataset.csv
 │
 ├── Project_Video.mp4
 │
 ├── README.md
 │
 ├── app.py
 │
 ├── config.yml
 │
 ├── credentials.yml
 │
 ├── domain.yml
 │
 ├── endpoints.yml
 │
 ├── requirments.txt
 │
 ├── train_models.py