Abstract In this research proposes a listeners dis-fluencies prediction system, this implementation suggests that stuttered speech identification and classification model. When the speaker produces a stuttered speech, a listener can affect the dis-fluencies of speech. In this work addresses the visual participant’s stuttered words and self-repairing classification model. The participants hear the sentences spoken by a speaker who stuttered and some heard sentences are spoken by the same speaker who produces the sentences without stuttering, in this scenario stuttered speech automatically repair by proposed MFLOR machine learning model. Results simulated has more accuracy compared to previous work; in this demonstration, the listeners can hear the processed speech. This work accomplished by applying the magnitude prediction technique and filtration on the select address. The accuracy 98.99%, sensitivity 92.34% and true positive rate 0.99 have increased, this work competes with existed methods.