Predicting Caesarean Section by Applying Nearest Neighbor Analysis

Maternal mortality and childbirth complications are major problem of delivery in rural area of many developing countries. In information era, it would be beneficial if the risk of delivery from uncertainty information could be informed or recommended to patients at earlier sign. As well as, physicians could draw approximate decision before it occurred.This paper proposes a modified nearest neighbor analysis, which is called CPD-NN algorithmto approximate risks about Caesarean sections due to Cephalopelvic disproportion (CPD). In the CPD-NN algorithm, it consists of three phases: initial phase, distance measure phase, and predicting phase. Herein, two determined distances are applied. First, the threshold distance, Dmin, is set to identify the closest neighbors. Dmaxis defined to identify the farthest neighbors. The k-neighbors, here, is dynamic, which is located within defined distances above. The results show that the efficiency and accuracy of CPD prediction are based on the number of training cases, dynamical k value, and similarity measures with different rules. Finally, the accuracy is 100% of predicting when applying the nearest rule in cosine similarity or correlation by 100 training cases with k ≈ 20, as well as 400 training cases with k ≈ 5