Learning embeddings of a heterogeneous behavior network for potential behavior prediction

Potential behavior prediction involves understanding the latent human behavior of specific groups,and can assist organizations in making strategic decisions.Progress in information technology has made it possible to acquire more and more data about human behavior.In this paper,we examine behavior data obtained in realworld scenarios as an information network composed of two types of objects(humans and actions)associated with various attributes and three types of relationships(human-human,human-action,and action-action),which we call the heterogeneous behavior network(HBN).To exploit the abundance and heterogeneity of the HBN,we propose a novel network embedding method,human-action-attribute-aware heterogeneous network embedding(a4 HNE),which jointly considers structural proximity,attribute resemblance,and heterogeneity fusion.Experiments on two real-world datasets show that this approach outperforms other similar methods on various heterogeneous information network mining tasks for potential behavior prediction.

Peripheral zone PSA density:a predominant variable to improve prostate cancer detection efficiency in men with PSA higher than 4 ng ml^(-1)

To improve the diagnostic efficiency of prostate cancer(PCa)and reduce unnecessary biopsies,we defined and analyzed the diagnostic efficiency of peripheral zone prostate-specific antigen(PSA)density(PZ-PSAD).Patients who underwent systematic 12-core prostate biopsies in Shanghai General Hospital(Shanghai,China)between January 2012 and January 2018 were retrospectively identified(n=529).Another group of patients with benign prostatic hyperplasia(n=100)were randomly preselected to obtain the PSA density of the non-PCa cohort(N-PSAD).Prostate volumes and transition zone volumes were measured using multiparameter magnetic resonance imaging(mpMRI)and were combined with PSA and N-PSAD to obtain the PZ-PSAD from a specific algorithm.Receiver operating characteristic(ROC)curve analysis was used to assess the PCa detection efficiency in patients stratified by PSA level,and the area under the ROC curve(AUC)of PZ-PSAD was higher than that of PSA,PSA density(PSAD),and transition zone PSA density(TZ-PSAD).PZ-PSAD could amend the diagnosis for more than half of the patients with inaccurate transrectal ultrasonography(TRUS)and mpMRI results.When TRUS and mpMRI findings were ambiguous to predict PCa(PIRADS score≤3),PZ-PSAD could increase the positive rate of biopsy from 21.7%to 54.7%,and help 63.8%(150/235)of patients avoid unnecessary prostate biopsy.In patients whose PSA was 4.0–10.0 ng ml^(−1),10.1–20.0 ng ml^(−1),and>20.0 ng ml^(−1),the ideal PZ-PSAD cut-off value for predicting clinically significant PCa was 0.019 ng ml^(−2),0.297 ng ml^(−2),and 1.180 ng ml^(−2),respectively(sensitivity>90%).Compared with PSA,PSAD,and TZ-PSAD,the efficiency of PZ-PSAD for predicting PCa is the highest,leading to fewer missed diagnoses and unnecessary biopsies.