Anomaly Detection Using Keras - Low Prediction Rate

Question

I built an Anomaly detection system using Autoencoder, implemented in keras. My input is a normalized vector with length 13. My dataset contains about 25,000 non anomaly inp