Global Monthly Temperature Mean Predictor using GRU Recurrent Neural Network (RNN)

Authors: Leo Martinez III - LinkedIn

Contact: leo.martinez@students.tamuk.edu

Created: Summer 2024

---

This Python program implements a Gated Recurrent Unit (GRU) Recurrent Neural Network (RNN) for predicting global monthly temperature means. The dataset utilized for training and testing the model is the Global Temperature Time Series, available here.

Model Overview:

• Language: Python 3.18
• IDE: Spyder
• Libraries Used:
  • numpy
  • pandas
  • matplotlib
  • sklearn
  • tensorflow

Implementation Details:

• Data Loading: The dataset is loaded and sorted by date.
• Data Visualization: The data is plotted to visualize the training and testing sets.
• Data Scaling: The data is scaled using MinMaxScaler to normalize the values.
• Data Splitting: The dataset is split into sequences of 250 steps for training the model.

Model Architecture:

• Input Layer: Takes sequences of 250 time steps with 1 feature.
• GRU Layer: Contains 125 units with 'tanh' activation.
• Dense Layer: Fully connected layer with 1 unit.

Training:

The model is compiled with the RMSprop optimizer and mean squared error loss function. It is trained for 50 epochs with a batch size of 32.

Evaluation:

The model's performance is evaluated on the test set, and predictions are made. The root mean squared error (RMSE) is calculated to measure the prediction accuracy.

Visualization:

The plot of the global temperature prediction is saved as Prediction_GRU.png with high DPI (400).

Note:

• Program was created in Spyder and separators were utilized ('#%%') specifically for Spyder IDE
• A similar project using the same dataset, but instead using LSTM can be found on my GitHub Page