Data is now a prime source of knowledge. The phrase, "The world's most valuable resource is no longer oil, but data," is gaining wide acceptance, and there is growing interest in effectively using AI-based data analysis. Unfortunately, analyzing large-scale data incurs excessive computation cost. While approximate methods are commonly employed to reduce the computation cost, they cannot yield exact results; they sacrifice accuracy to improve efficiency. This talk will introduce representative methods of pruning unnecessary computations to achieve fast and accurate data analysis with reasonable cost.