Cleansing Noisy City Names in Spatial Data Mining

One of the biggest adversaries to data mining from a large data warehouse is poor data quality. It is because most data mining algorithms have been designed based on the assumption that the data is clean and meaningful. Hence, poor data quality may lead to completely unexpected results. In this paper, an automatic city name correction algorithm is proposed to cleanse a large spatial database without requiring human intervention or a prior knowledge of the context. The algorithm achieves a precision of 96.6% which is significantly better than the 86.6% of the traditional Levenshtein distance and the 92% of the Longest Common Subsequence algorithm.