Mini Project

Univariate time series forecasting

Mini project is based on univariate time series forecasting where we predic the milk production in a month based on data of milk productions in previous 3 months.

Our project is multivariate time series forecasting as it takes multiple features as input and predicts the closing price.

Hence, to form a basic understanding we decided to use a single variable time series forecasting as our mini project.

How it works

• Data collection and preprocessing by using min-max scaler since LSTM is senstive to scale of input data
• Defining sequence : Sliding Window Approach where each input sequence consists of a fixed number of past time steps
• Splitting the data : Make train,validation and test set, take a sequence length of 3
• Building the model : Use an sequential model with Input, LSTM, Dense and Dropout layers
• Training model : Use adam optimizer to adjust parameters and MSE for loss
• Finally run the fit method to run the model on training set
• Make predictions on the test data and plot the graph

Loss graph

Predicted vs Actual