A General Suspiciousness Metric for Dense Blocks in Multimodal Data

Detecting deceptive engagement in social media by temporal pattern analysis of user behaviors: a survey

“…Among them even fewer are practical for large‐scale social media platforms. This is also pointed out in the recent survey by Yu et al As a result of this search, we selected eleven recent papers . The approaches in these papers have been tested or are actively in use in real social network media platforms.…”

Fraud Detection in Comparison-Shopping Services: Patterns and Anomalies in User Click Behaviors

“…4. It is also a good criterion of anomalousness to examine whether a user produces high density in a block [8]. Here, the density in a block indicates not only the ratio of non-zero numbers but also the magnitude of numbers in the block.…”

“…SVD of an m × n matrix A is a factorization of the form A = UΣV T : U and V are m × m and n × n matrices whose columns are called the left-singular vectors and right-singular vectors of A, respectively; Σ is an m×n diagonal matrix comprised of its singular values. Here, the top singular values and singular vectors indicate dense blocks in the matrix [8]. Also, for each top leftsingular vector, the m absolute values, which are referred to as eigenscores, indicate the degree of relevance of each user to the corresponding dense block: the higher the value, the more engaged in the corresponding dense block [8].…”

“…Here, the top singular values and singular vectors indicate dense blocks in the matrix [8]. Also, for each top leftsingular vector, the m absolute values, which are referred to as eigenscores, indicate the degree of relevance of each user to the corresponding dense block: the higher the value, the more engaged in the corresponding dense block [8].…”

“…Then, we find dense blocks in the matrix and measure how much relevant each user is to the found dense blocks. Towards this goal, we employ the singular value decomposition (SVD), one of the most widely used methods to discover such dense blocks [8]. SVD of an m × n matrix A is a factorization of the form A = UΣV T : U and V are m × m and n × n matrices whose columns are called the left-singular vectors and right-singular vectors of A, respectively; Σ is an m×n diagonal matrix comprised of its singular values.…”