LAUSR

Abstract Nowadays, NAND flash is widely used for its excellent characteristics. However, the increasing storage capability leads to the decrease of reliability of NAND flash. Therefore, improving the reliability of NAND flash chip has become a hot issue to be solved. If we can find an effective method to predict the error rate distribution of NAND flash, it will provide significant guidance for adapting appropriate error correction algorithm and wear-leveling algorithm. Based on the 200 days of measured data from NAND flash experimental platform, a comprehensive error rate prediction model of NAND flash is established using the random forest algorithm, and the experimental results of the model are analysed and evaluated. And the results show that the proposed stochastic forest algorithm not only has high prediction accuracy, but also operates fast.