Assignment: Machine Learning Data Web Application

Overview

The task in this assignment is to develop a web application storing data that can be used for a machine learning model.

Data stored by the web application are a collection of data points that involve:

• Two or more continuous features
• A single categorical feature

The exact names and the nature of these features do not matter and can be chosen arbitrarily. The data are supposed to be used for a classification task performed by a machine learning model that uses the continuous features to predict the corresponding category. The web application stores the data in a relational database consisting of a single table that holds individual data points as records.

Versions of the Application

• Basic version: Data points can be retrieved from, added to, and deleted from the database.
• Extended version: It is also possible to make predictions of the most likely category based on submitted values of the continuous features using a machine learning model.

The web application consists of two parts: a website and an API. The website involves HTML pages to be rendered in a browser, whereas the API involves data transfer directly over HTTP.

---

Requirements for the maximum grade of 4

Using a web framework (such as Flask, Django, FastAPI, etc.) and a database management system (such as SQLite, PostgreSQL, MySQL, MariaDB, etc.), implement the website part of the basic version of the web application.

Database Specification

The database should consist of a single table with the following columns:

• Primary key: represented as integers.
• Continuous features: two or more columns (one per feature, represented as floating-point numbers).
• Categorical feature: one column (represented as integers).

Website Paths

• / – Home page: displays all data points in a table. The table should include sequence numbers, columns for each continuous feature, and a column for the categorical feature. It should also allow the user to invoke deletion of a record via the /delete/<record_id> path.
• /add – Add data: contains an HTML form for a new data point.
  • Submitted via HTTP POST.
  • Validation: If successful, add record and redirect to home. If failed, return 400 HTTP status and an error page.
• /delete/<record_id> – Delete data:
  • Accepts only HTTP POST requests.
  • Validation: Check if record exists. If successful, delete and redirect to home. If failed, return 404 HTTP status and an error page.

API Endpoints

• GET /api/data – Returns all data points as a JSON list of dictionaries.
• POST /api/data – Adds a new data point via JSON. Returns the primary key of the new record or a 400 error with a message.
• DELETE /api/data/<record_id> – Deletes a data point. Returns the primary key of the deleted record or a 404 error.

Note: Use Object-Relational Mapping (ORM) (e.g., SQLAlchemy, Peewee) for both website and API parts.

---

Requirements for the maximum grade of 5

All requirements for the grade of 4 must be satisfied.

Extended Website Part

• /predict – Predicts a category based on features.
  • Contains an HTML form for continuous features.
  • Submitted via HTTP POST.
  • Success: Feed values to the ML model and display the predicted category.
  • Failure: Return 400 HTTP status and an error page.

Machine Learning Model

• Classifier: k-nearest neighbors (k-NN).
• Parameter: $k$ parameter should not be larger than 5 ($k \le 5$).
• Training: Trained on all data points currently in the database.
• Preprocessing: Each continuous feature must be standardized before training and before prediction.
• Package: Use scikit-learn.

Extended API Part

• GET /api/predictions – Predicts a category.
  • Values are passed via query parameters.
  • Success: Returns JSON with the predicted category.
  • Failure: Return 400 HTTP status and a JSON error message.

---

It is recommended to use the Requests package for testing the API.