- The dataset contains four main features: Hour, Lane 1 Flow (Veh/Hour), Lane 1 Occ (%), and Lane 1 Speed (mph).
Visualizations of Traffic Features Over Time:
- Data Normalization using MinMaxScaler:
Visual representation showcasing the process of normalizing the data.
The build_lstm_model function defines a sequential model, incorporating LSTM layers that allow the network to retain memory over sequences, enabling the model to capture temporal dependencies within the traffic data.
Role of input_shape and activation function:
The code specifies the input shape and activation function within the LSTM layer. The input_shape=(seq_length, num_features) denotes the input data shape passed to the LSTM layer, indicating the sequence length and the number of features. Moreover, the activation function used, in this case, relu (Rectified Linear Unit), helps introduce non- linearity within the LSTM layer, allowing the model to learn complex patterns and relationships within the traffic data. model.compile(optimizer='adam', loss='mean_squared_error')
This line specifies the Adam optimizer to be
used in the compilation of the LSTM model, employing the mean squared error as the loss function The line model.fit(X_train_short, y_train_short, epochs=epoch, batch_size=64) implies a batch size of 64. Therefore, the model is being trained using mini-batch gradient descent, where the dataset is divided into smaller batches, and the model parameters are updated after processing each batch. This strategy combines the advantages of stochastic gradient descent (updates after every sample) and full batch gradient descent (updates after the entire dataset) to achieve a balance between efficiency and robustness in training. Default hyper-parameter values such as epochs and batch size are used to train initial models for short and long-term predictions. The resulting Root Mean Square Error (RMSE) values are calculated and visualized to understand the accuracy of these predictions, helping assess the model's performance. Finally, a comparative graph displaying the RMSE against the varying epoch values is created, providing a comprehensive view of the impact of different epochs on prediction accuracy. Analyzing how varying epochs affect both long-term and short-term predictions sheds light on the intricate connection between training duration and the accuracy of predictions. This examination offers valuable insights into determining the most effective training durations for each specific forecasting horizon within LSTM models. a subplot visualization of epochs versus short-term predictions a subplot visualization of epochs versus Long -term predictions Comparison between Short-Term Predictions and Actual Short-Term Values: Comparison between Long-Term Predictions and Actual Long-Term Values: Thank You