A survey of neural network-based cancer prediction models from microarray data

Highlights：

• The survey is distinguished from previous works by analysing the contributions that included three basic components which are: Neural networks, gene expression datasets and cancer prediction. The survey also presents technical details related to data preprocessing, model configuration, learning parameters and the evaluation metrics which give the reader ideas about recent approaches.

• We grouped the considered works according to the neural network functionality in the model. Starting with presenting preprocessing techniques, then the models configuration and evaluation metrics and a summary at the end.

• A discussion for better future practice was given at the end to highlight some practical issues that can be considered to increase future models predictability.

摘要

•This survey paper presents recent works proposing neural network models for cancer prediction related problems. All the considered works used gene expression datasets to experiment the proposed models.•The survey is distinguished from previous works by analysing the contributions that included three basic components which are: Neural networks, gene expression datasets and cancer prediction. The survey also presents technical details related to data preprocessing, model configuration, learning parameters and the evaluation metrics which give the reader ideas about recent approaches.•We grouped the considered works according to the neural network functionality in the model. Starting with presenting preprocessing techniques, then the models configuration and evaluation metrics and a summary at the end.•A discussion for better future practice was given at the end to highlight some practical issues that can be considered to increase future models predictability.