Abstract: Alzheimer's disease (AD) is a neurodegenerative disease that seriously threatens the life of the elderly and there is no effective therapy to treat or delay the onset of this disease. Due to the multifactorial etiology of this disease, the multi-target-directed ligand (MTDL) approach is an innovative and promising method in search for new drugs against AD. In order to find potential multi-target anti-AD drugs through reposition of current drugs, the database of global drugs on market were mined by an anti-AD multi-target prediction platform established in our laboratory. As a result, inositol nicotinate, cyproheptadine, curcumin, rosiglitazone, demecarium, oxybenzone, agomelatine, codeine, imipramine, dyclonine, melatonin, perospirone, and bufexamac were predicted to act on at least one anti-AD drug target yet act against AD through various mechanisms. The compound-target network was built using the Cytoscape. The prediction was validated by molecular docking between agomelatine and its multiple targets, including ADORA2A, ACHE, BACE1, PTGS2, MAOB, SIGMAR1 and ESR1. Agomelatine was shown to be able to act on all the targets above. In conclusion, the potential drugs for anti-AD therapy in the database for global drugs on market was partially uncovered using machine learning, network pharmacology, and molecular docking methods. This study provides important information for drug reposition in anti-AD therapy.