HYBRID SINE COSINE AND FITNESS DEPENDENT OPTIMIZER FOR INCOMPLETE DATASET

The hybrid sine cosine and fitness dependent optimizer (SC-FDO) introduces four modifications to the original fitness dependent optimizer (FDO) algorithm to improve its exploit-explore tradeoff with a faster convergence speed. The modifications include a modified pace-updating equation, a random weight factor and global fitness weight strategy, a conversion parameter strategy, and a best solution-updating strategy. This chapter evaluates the generalization ability of the hybrid SC-FDO-based neural network (SC-FDONN) in handling missing data imputation challenges that exhibit different percentages of missingness. The hybrid SC-FDONN's performance was evaluated using hold-out and cross-validation techniques. The findings revealed that the SC-FDONN outperformed all the benchmarks by an average accuracy of 94.3%. Therefore, the hybrid optimizer, SC-FDONN, is an effective technique for handling different percentages of missing data problems.