The invention relates to a multidimensional data based detecting method of traffic abnormal spots. According to the method, for flow, speed and lane occupancy ratio data of each section, obtained by a microwave apparatus in a continuous period, historical jam probability of each section is calculated according to the speed data, positive and negative anomalies of each section are defined through comparing recent traffic state index values, anomaly degree is calculated for each section with the negative anomaly by means of a density-based local anomaly factor method, and weighted anomaly degrees are calculated according to positive and negative anomaly factors and are ranked. The method has the advantages that multi-index data is utilized, the uniformity in sample spatial data distribution is considered, local limitedness of the density-based local anomaly factor method is avoided the use of the features of traffic data, road abnormal spots can be effectively detected, the traffic administration is helped command the road traffic, service efficiency of roads is adjusted and optimized, and the method features high universality, high feasibility, high reliability and high applicability.