The proliferation of data-driven methods in financial analytics has underscored the importance of effectively handling missing data and accurately forecasting market behaviors. This review consolidates recent advancements in missing data imputation within financial datasets, alongside machine learning (ML) and statistical models applied to cryptocurrency forecasting. We highlight classical approaches, such as expectation-maximization and k-nearest neighbors, and explore contemporary frameworks involving deep learning architectures like LSTM and generative adversarial networks (GANs). Additionally, we compare hybrid optimization methods tailored to the volatility of cryptocurrency markets. This study synthesizes key findings, identifies prevailing challenges, and outlines promising directions for future research in financial data analytics.