算法分析读后

A Microeconomic Data Mining Problem:Customer-Oriented Catalog Segmentation
本文讨论了greedy算法的不足，只能找到局部极值，因为牵扯到the heuristic nature of the quality criteron，而又没有回顾backtracking (after sub-optimization或前瞻look ahead能力

改进办法适用randomize algorithm in a two-step approach(Random-Product-Fit):

1. Use greedy to find segments

2. resulting catalogs and corresponding clusters are iteratively optimized by randomlyre placing one catalog product bya non-catalog product (Random-Product-Switch).