Outlier detection was instigated as noise removal technique for enhancing the prediction accuracy of Machine Learning (ML) algorithms. In due course, outlier detection emerged as the phenomenon of mining rare/alarming patterns to assist in decision making process. Outliers can be point or collective types and can be detected by the supervised and unsupervised ML algorithms. Such algorithms have the probability of missing out certain point or collective outliers in high dimensional quantitative data. To overcome the aforementioned issue, this work proposes a Semi Supervised Outlier Detection (SSOD) algorithm with a probability based surprising measure ‘Lift’ for outlier detection. The performance proposed SSOD is benchmarked with the existing outlier detection ML algorithms. By studying performance of the algorithms, it is understood that the proposed SSOD with Lift measure outperforms the benchmarked ML algorithms.