Student Dropout Prediction for University with High Precision and Recall

Since a high dropout rate for university students is significant risk to local communities and countries, prediction model using machine learning an active research domain prevent from dropping out. However, it challenging fulfill the needs of consulting institutes office academic affairs. To institute, accuracy in utmost importance; offices affairs other offices, reason out essential. This paper proposes Student Dropout Prediction (SDP) system, hybrid predict who are about drop university. The tries increase precision recall predicting dropouts. We then analyzed by compressing feature set with PCA applying K-means clustering compressed set. SDP system showed value 0.963, which 0.093 higher than highest-precision existing works. F1 scores, 0.766 0.808, respectively, were also better those gradient boosting 0.117 0.011, making them highest among works; Then, we classified reasons into four categories: “Employed”, “Did Not Register”, “Personal Issue”, “Admitted Other University.” University” was highest, at 0.672. In post-verification, increased counseling efficiency accurately dropouts “High-Risk” group while including more total addition, presenting guidelines each department, could receive personalized counseling.