본문 바로가기

컴퓨터 & 프로그래밍/My Papers

Enhanced Back Propagation Algorithm for Estimating Ecological Data with Missing Values

[IJ 3] Won-Du Chang, Jungpil Shin, Mi-Young Song, Tae-Soo Chon, "Enhanced Back Propagation Algorithm for Estimating Ecological Data with Missing Values", WSEAS Transactions on Computers, Issue 9, vol. 5, pp.2043-2048, 2006. [pdf]

Abstract: Multi layer perceptron with back propagation algorithm is popular in various fields of investigation as a non-linear predictor. Though MLP can solve complex and non-linear estimation problems, it cannot use incomplete data patterns for training directly. The estimation of macroinvertbrates is important issue in ecological informatics, and data missing occurs frequently and inevitable in many cases, which may lead to a biased estimation. We propose a training algorithm for the data including missing values, using conventional MLP network. Focusing on the fact that BP algorithm uses the amount of the error and its sign to modify the weights, we redefined the activation function using the minimum error for incomplete pattern and stopping criterion. Proposed algorithm is compared to ignoring and MLP replacing algorithms using polinomial data. Through the polinomial experiment, it could learn from incomplete patterns and avoid biased learning from misestimating of missing values successfully showing heigher correlation and lower RMSE than other algorithms. It is also applied to real ecological data. In the estimation problem of Chironomus flaviplumus richiness, it increased the estimation correctness using incomplete data patterns.