Lossless Data Compression Method Using Deep Learning

A method of lossless data compression on various data formats using predefined algorithms and machine learning is claimed in this chapter. The proposed system makes use of predefined algorithms and a neural network for encoding and decoding. As it uses machine learning (sequence to sequence) model, it makes the system adaptive and the compression ratio keeps improving over time with more and more data. Use of machine learning ensures that different file formats can be compressed within the same system. Compressed data obtained after the implementation of predefined algorithms will be used as output labels for the ML model, and the original data will be the input labels for the ML model. For decompression, the compressed data is used as input, and the original data is used as output labels. The system can be deployed as a cloud-based application in which the user can input data and store it in a cloud database. The proposed system achieves 74% accuracy on a batch comprising 50,000 training data. Higher accuracy can be achieved by training the model on more data.