Research on non-time-series data filling methods based on feature evaluation

Abstract With the rapid development of information age, a large amount data is used in popular research areas such as mining. Missing has very serious impact on both process and result mining, it important to find out how fill missing values accurately efficiently. In this paper, we propose method optimally based backpropagation evaluation functions for non-time-series data. Based target value error its own after filling, four classical filling methods, namely mean, interpolation, model prediction, K-nearest neighbor, are considered selection. Finally, single-model padding multi-model weighted schemes compared, results show that with highest fitness selected work best different degrees missingness datasets.