Skip to content

Latest commit

 

History

History
89 lines (75 loc) · 4.24 KB

README.md

File metadata and controls

89 lines (75 loc) · 4.24 KB
Raw

Would You Survive The Titanic --version 2.0

  • This is a Flask API developed by me to determine if an onboarder on the infamous Titanic would survive provided their details.
  • This version is built upon the previous version which was quite basic in operation.

  • New features include:

    • New models and model selection options
    • New features for more accurate prediction
    • New form format

  • Contents:

Deploy the API on Heroku by clicking the button below.

Deploy

❗ Some models are not working currently. Specifically: Random Forest, KNN and Decision Trees. The solution is being worked on.

Running the project locally:

- Clone this repository. Open CMD. Ensure that you are in the project home directory. Create the machine learning model by running models.py as such:

python models.py

- This will create a serialized version of our models into files with an extension .pkl or use the previously pretrained models saved in the ./models folder.
- Now, run api.py using below command to start Flask API

python api.py

- Open any browser and paste this URL: http://localhost:5000 to run the file as an app.

Following images show how the API will look when run properly:

Dataset:

Variable Definition Key
pclass Ticket class 1 = 1st, 2 = 2nd, 3 = 3rd
sex Sex
Age Age in years
sibsp # of siblings / spouses aboard the Titanic
parch # of parents / children aboard the Titanic
ticket Ticket number
fare Passenger fare
cabin Cabin number
embarked Port of Embarkation C = Cherbourg, Q = Queenstown, S = Southampton

  • Target:

    • 0 : Not Survived,
    • 1 : Survived

Models Used:

  • The dataset shows that this is clearly a classification task and can be solved by a myriad of classification algorithms such as Logistic Regression, Decision Trees and even Random Forests.
  • I chose 6 algorithms to train the dataset on because why not.
  • The models which were selected were: Logistic Regression, K-Nearest Neighbours, Gaussian Naive Bayes, Decision trees, Random Forest and Support Vector Machines.
  • Model Performances:
Model Score
Random Forest 86.76
Decision Tree 86.76
KNN 84.74
Logistic Regression 80.36
Support Vector Machines 78.23
Naive Bayes 72.28

API:

  • I made an API for all the models so that users can interact and use the Machine Learning models with ease. User can select which model they would like to use during prediction.
  • To make the API work I have used the Flask library which are mostly used for such tasks.
  • I have also connected a HTML form to the flask app to take in user input and a CSS file to decorate it.

Deployment:

  • The Flask API was deployed on the Heroku cloud platform so that anyone with the link to the app can access it online.
  • I have connected this GitHub repository to Heroku so that it can be run on the Heroku dyno.
  • I have used the Gunicorn package which lets Python applications run on any web server. The Procfile and requirements.txt should be defined with all the details required before the deployment.

What did I learn:

  • Data Wrangling using Pandas
  • Feature Engineering to fit our data to our model
  • Saving the model and using it again with Pickle
  • Making a flask app
  • A little frontend web development
  • Making the app live by deploying it on cloud platforms

Back to Top