PREDIKSI PENGEMBALIAN ATAU RETUR DALAM BELANJA ONLINE MENGGUNAKAN METODE KLASIFIKASI K-NN (KNEAREST NEIGHBOR); PREDICTION OF RETURN SHIPMENT IN ONLINE SHOPPING WITH CLASSIFICATION USING K-NN (K-NEAREST NEIGHBOR) METHOD

Roqimawati, Astika P, Aina Musdholifah

2015 | Skripsi | FMIPA

Shopping on the internet has become an alternative for customers, but it presents an obstacle with the product which are not in accordance with customer expectations. To overcome this, many online shops provide free return policy. Because the cost of return shipment are borne by the online shop, therefore shop have to predict returns of customers who have made purchases. With this prediction, the online shop is expected to prepare a solution to prevent returns on similiar transaction will not happen again. K-Nearest Neighbor algortihm or k-NN is a classification algorithm that generates strong data and effective when used on large data. By using k-NN algorithm that applies to the object classification method based on learning data, predictions can be sought from the similarity according to the number of nearest neighbors or the value of k. The process of k-NN algorithm in this study begins with a parameter optimization to select the optimal values of k, the training phase, the phase of classification, and take as many as k data closest to seek a majority label. Testing the system in this study can be predict the return shipment for the past and the future. In testing the prediction of return shipment system generate the most optimal value of k is k = 21 with F-measure value 0.6427. Accuracy testing of the historical purchase data of online shop in 2013 amounted to 58.94%.

Kata Kunci : prediction; return shipment; online shopping; k-Nearest Neighbor algorithm, k-NN