Detection of Spatial Outlier by Using Improved Z-Score Test

Outlier detection is a prominent technique in data mining since it has various vital applications. Outlier detection can be used to remove noise or study the specific observations in data that are dissimilar to those in its neighborhood. The detection of outliers are being used to show out hidden and valuable information in many areas of applications. [1]. The z-score test has long been used to detect outliers in data. In this work, a modified version of the z-score test is proposed which can lead to a reduction of the time-complexity of the traditional z-score test by a value of n each iteration, where n donates the number of spatial points in the data set which have not been discovered as outliers. Using the outlier threshold value to calculate the value of the modified z-score function rather than calculating the z-score value for each observation in the dataset makes this possible.