Depression is a serious medical illness that is caused by various causes. In this study, we have developed prediction models by classifying depression datasets which are taken from Kaggle. We have focused on feature selection. After preprocessing data, we have selected features by applying Logistic Regression, Decision Tree and Correlation Matrix methods. Those feature engineering techniques are helped in finding effective and best features from the dataset. A large number of instances is used to our study for developing prediction models. Then we have applied different algorithms such as Logistic Regression, K-NN, SVM, Naive Bayes for classifying data and building models. We have gotten the best classification model and accuracy of K-NN which is 79%. The applied Logistic Regression, SVM, and Naïve Bayes have 77% accuracy, but K-NN showed the best Accuracy.