This repository contains a solution notebook for an interview screening project conducted for an energy trading company. The primary objective of the project was to develop an algorithm capable of forecasting electricity prices for the next 24 hours, starting from 00:00 of the following day, and to do it consequently day by day. It's important to note that the dataset and the task description provided by the company are proprietary and cannot be included in this repository.
Project Overview The project involved analyzing hourly power prices spanning from January to April. The main focus was on accurately predicting April's prices. The notebook is structured into five key stages:
Data Loading and Inspection: An initial exploration of the dataset was conducted to understand its structure and integrity.
Target Variable Analysis: Various visualizations and statistical tests were employed to gain insights into the characteristics of the target variable.
Missing Values Handling: Strategies were developed to identify and handle missing data effectively.
Feature Engineering: Existing variables were optimized, and new features were created based on the target variable to enhance predictive performance.
Predictive Modeling: A predictive model was constructed and fine-tuned to forecast April's hourly power prices.
Please note that only the code developed as part of this project is included in this repository. Any proprietary data or task descriptions provided by the company have been excluded.