Integration of deep feature extraction and ensemble learning for outlier detection

Evaluation of Unsupervised Anomaly Detection Methods in Sentiment Mining

“…Hesam Izakian et al, (2013) in their work they demonstrated that a huge piece of the accomplishments in anomaly detection because of the novel achieved in distance measurements and dimensionality reduction of time series data. In any case, tending to the matter of time arrangement grouping through ordinary methodology has not explained the issue totally, particularly if the class labels of time series are indistinct [6]. Sophisticated algorithms don't just mark perceptions as anomaly, yet relegate scores to perceptions, speaking to degrees or probabilities of outlierness.…”

“…Sophisticated algorithms don't just mark perceptions as anomaly, yet relegate scores to perceptions, speaking to degrees or probabilities of outlierness. Some famous models depend on the distance between objects [3][4][5][6][7][8], or based on the variance of angles between object vectors [2,6,7,9] or on the density of the neighborhood of an object [13,19], or on other principles of outlierness in various domains [12,13]. Wang et al, [20] proposed an outlier detection method based on clustering.…”

“…Later, it employs a heuristic to group the resultant clusters into huge and undersized clusters. In the end, an anomaly value is calculated using the distance of every instance to its cluster centre multiplied by the data instances belonging to its cluster [1,6]. In case of clusters of smaller size, the distance to the closest large cluster is used.…”

“…In data mining, the identification of "abnormal" instances in the dataset is an emerging field incredible interest. Detection of abnormal instances is generally said to be as "anomaly detection" [6]. Anomalies are not the same as the normal instances with reference to their attributes and they are unusual in the dataset with respect to regular instances [1,2,5].…”

“…It has an extensive count of users distributed throughout the world and results in massive volume of social media data [7,8,11]. Analyzing such social media data to detect abnormal instances in a timely scale is a valuable topic of research [2,6,10]. In this study, a comparative evaluation of various anomaly detection algorithms such as density based, statistical based and cluster based anomaly detection methods is presented.…”