Hemodialysis　is　one　of　the　treatments　for　patients　with　end-stage　renal　disease.　The　quality　of　dialysate　scheme　directly　affects　the　prognosis　of　dialysis　patients.　In　order　to　achieve　the　goal　of　making　individualized　dialysate　scheme　for　dialysis　patients,　improve　the　decision-making　efficiency　of　doctors　and　reduce　the　decision-making　pressure,　a　dialysate　decision-making　model　based　on　deep　learning　was　studied　and　proposed.　In　this　model,　the　bidirectional　long　short-term　memory　(BLSTM)　is　used　to　study　the　multi-dimensional　temporal　physiological　records　of　dialysis　patients　in　both　positive　and　negative　directions,　capturing　the　physiological　characteristic　information　of　patients.　Besides,　we　introduce　the　attention　mechanism,　in　order　to　capture　the　important　information　of　time　points,　which　enhances　the　interpretability　of　the　model.　Experiments　show　that　this　model　has　higher　macro　precision,　macro　recall　and　macro　F1　than　other　models.　©　2020　IEEE.