Merged ETT Dataset for Time Series Forecasting

Instructions for Using the Merged ETT Dataset:

1. Download the Dataset
   Retrieve the merged_ETT_dataset.csv file along with individual files (ETTh1.csv, ETTh2.csv, ETTm1.csv, ETTm2.csv) if you want to compare or analyze separately.

2. Load the Dataset
   Use Python with pandas to load the dataset:

   import pandas as pd df = pd.read_csv("merged_ETT_dataset.csv", parse_dates=["date"]) df.set_index("date", inplace=True)

3. Understand the Features

   • HUFL, HULL: High-usage frequency limits (upper/lower)

   • MUFL, MULL: Medium-usage frequency limits (upper/lower)

   • OT: Operational Temperature (target variable)

4. Preprocessing Suggestions

   • Handle missing values if any (e.g., df.dropna()).

   • Normalize the dataset using standard techniques (MinMaxScaler, StandardScaler).

   • For forecasting tasks, create windowed sequences of past n timestamps as input.

5. Modeling Guidelines

   • For statistical methods: Use ARIMA, Prophet, or Holt-Winters.

   • For deep learning: Suitable for LSTM, GRU, Transformer, Informer, Autoformer, etc.

   • Define the OT column as the target variable for prediction.

6. Use Cases

   • Long-term electricity demand forecasting

   • Anomaly detection in transformer operations

   • Predictive maintenance scheduling

   • Smart grid optimization and planning

7. Visualization

   • Visualize time series trends using:

     df["OT"].plot(figsize=(15, 4), title="Operational Temperature Over Time")

8. Citing the Dataset

   • Please cite the IEEE DataPort page or corresponding data descriptor article if you use this dataset in your research or applications.