Long short-term memory (LSTM) RNN in Tensorflow

Long Short-Term Memory (LSTM) where designed to address the vanishing gradient issue faced by traditional RNNs in learning from long-term dependencies in sequential data. LSTMs are capable of maintaining information over extended periods because of memory cells and gating mechanisms. These memory cells are managed by three primary gates: the input gate, the forget gate and the output gate.

1. Importing Libraries

In this step, we will import the necessary libraries like pandas, numpy, matplotlib, scikit-learn and tensorflow. Here tensorflow library is used to create the LSTM Model.

2. Data Loading, Preparing and Scaling

Here we are using a dataset of monthly milk production using LSTM. You can download dataset from here.

• We load the dataset of monthly milk production. The "Date" column is converted to datetime format for time series analysis.
• We scale the data to a range of [0, 1] using MinMaxScaler to help the model train more effectively.

3. Creating Sequences and Train-Test Split

Here we generate sequences of input data and split the dataset into training and testing sets.

• We use a sliding window of 12 months (1 year) of past data to predict the next month's production.
• The dataset is split into training and testing sets and reshaped to match the LSTM input shape.
• We split 80% data for training and 20% for testing purposes.

4. Building the LSTM Model

This step involves defining and building the LSTM model architecture.

• The model consists of two LSTM layers, each with 128 units and a dropout layer after each to prevent overfitting.
• The model concludes with a Dense layer to predict a single value (next month's production).

5. Training and Evaluating the Model

In this step, we train the model on the training data and evaluate its performance.

• The model is trained for 100 epochs using a batch size of 32, with 10% of the training data used for validation.
• After training the model is used to make predictions on the test set and we calculate the Root Mean Squared Error (RMSE) to evaluate performance.

Output:

Model trains for 100 epochs with a batch size of 32, using 10% of the training data for validation.

6. Visualizing Model Performance

In this step, we visualize the actual vs predicted values. A plot is generated to compare the actual milk production against the predicted values, allowing us to evaluate how well the model performs over time.

Output:

The LSTM model successfully captures the trends and patterns in the time series data. As observed, the predicted values closely follow the actual values with small variations during transitions between peaks and lows. This demonstrates the effectiveness of LSTM for time series prediction tasks such as forecasting milk production.

You can download source code from here.

You can implement LSTM using PyTorch also: Long Short Term Memory (LSTM) Networks using PyTorch