/time_series_final_project_data_analysis

It was the final project for the data analysis course my team and I compered among the performance of transformers, linear models and LSTM on the time series forecasting

Primary LanguageJupyter Notebook

Problem statement

Our problem is to investigate the effectiveness of Transformer based models for time series forecasting and compare their performance with LSTM and linear models named LTSF Linear, such as Linear. the patchTST paper: https://arxiv.org/abs/2211.14730 and linear models and FEDformer paper: https://arxiv.org/abs/2205.13504

Data

There are five datasets were used four of them were electrical power consumption time series from collected from industrial sites. Each data point consists of 8 features, including the date of the point, the predictive value ”oil temperature”, and 6 different types of external power load features. The features are “date”, ”HUFL”, ”HULL”, ”MUFL”, ”MULL”, ”LUFL”, ”LULL” and ”OT”.

The another dataset (Hourly energy demand generation and weather) contains 4 years of electrical consumption, generation, pricing, and weather data for Spain. Consumption and generation data was retrieved from ENTSOE a public portal for Transmission Service Operator (TSO) data. Settlement prices were obtained from the Spanish TSO Red Electric España.

models

Patch-TST

Linear-LSTF

FEDformer

LSTM

results

Screenshot (40)