Power Quality Disturbances Data Dimensionality Reduction Using Autoencoder

Dimensionality reduction is an important techniques to reduce the size of the data in terms of number of input features so that light weight models can be developed. In this paper, power quality disturbances dataset consists of 750 rows and 72 columns is reduced using autoencoderAutoencoder without loss of much information. Lower dimensional data consists of total 750 rows and 21 columns. It means that any machine learning model can be designed with 21 input features instead of 72 input features that leads reduction in complexity of model in terms model parameters. AutoencoderAutoencoder model is developed using python, from the results, it is observed that the autoencoderAutoencoder is compressing the original data with less validation loss and without much variance deviation from the original data.