The Bit Flipping Criteria Models based on Supervised Learning in Magnetic Recording System

The 2D interference still be the major problem for cancelation in the high areal density magnetic recording, such as in bit-patterned media recording (BPMR) systems. The two-dimensional modulation codes have been proposed to cancel the 2D interference effect, efficiently improving the overall system performance. However, the bit-flipping technique in the rate-2/3 modulation coded system can effectively flip the ambiguous data bits, it is difficult to confirm the optimal threshold value for giving the gain to flip those ambiguous data, especially while the systems face several noises. To improve this problem; therefore, we propose the simple 2/3 2D modulation code that provides the favorable condition for flipping through the criteria of bit flipping models based on several supervised learning in data science techniques, as Generalized Linear Model, Decision Tree, Deep Learning, Large Fast Margin, Gradient Boosted Trees, Random Forest, Naive Bayes, Logistic Regression and Support Vector Machine. These results found the high accuracy percentage of all algorithms in the low SNR level, especially in the Support Vector Machine model.