Outlier detection is an important branch in the areaof data mining,It has been widely used in weather forecasting, network intrusion detection, telecommunications and credit card fraud detection,etc. LOF algorithm has good detection effect and availability, but its computation is very high, whose efficiency is not good enough,And when calculating the distance between two objects, LOF algorithm ignores the different influence of different properties.To solve above disadvantages, we put forward an improved outlier detection algorithmiLOF^*. iLOF^* algorithm usesgrid to reduce the data sets, so as to improve the efficiency of the algorithm; at the same time, when calculating the distance between the object, iLOF^* algorithm gives different weights to different properties through the introduction of information entropy, which improve the accuracy of the algorithm.In addition, we use the parallel computing framework MapReduce to parallel iLOF * algorithm, which further improves the efficiency of algorithm on large data sets.The experimental results demonstrate the effectiveness and efficiency of the proposed algorithm.