Data Mining for Generating Predictive Models of Local Hydrology

摘要: The problem of downscaling the effects global scale climate variability into predictions local hydrology has important implications for water resource management. Our research aims to identify predictive relationships that can be used integrate solar and ocean-atmospheric conditions forecasts regional flows. In recent work we have developed an induction technique called second-order table compression, in which learning viewed as a process transforms consisting training data (which sets atomic values entries) with fewer rows by merging consistency preserving ways. Here, apply compression generate models future inflows Lake Okeechobee, primary source supply south Florida. We also describe SORCER, system compare its performance three well-established mining techniques: neural networks, decision tree associational rule mining. SORCER gives more accurate results, on average, than other methods average accuracy between 49% 56% prediction discretized four ranges. discuss these results practical issues assessing from guide decision-making.