This paper presents a data-driven prediction method for electric battery condition monitoring with different loads. The prediction is subject to the usage time of the electric battery during the discharge condition. Two variables are selected as the prediction input i.e. load (Watt) and discharge voltage (Volt) to predict how much time (hour) has left during the discharge period. Adaptive Neuro-Fuzzy Inference System (ANFIS) serves dual purposes of classification and prediction of battery discharge conditions, while Support Vector Machine (SVM) is implemented for classification comparison. While SVM demonstrates superior classification performance with 95% accuracy compared to ANFIS's 88%, ANFIS provides the added value of precise time prediction. The time-series data was collected from the discharge battery experiment for a few hours that uses the electric rechargeable battery from a fully charged capacity to an empty capacity. The experiments were conducted on four different load conditions i.e. 130, 180, 200, and 220 Watts. The prediction result of ANFIS was compared with the result of the Support Vector Machine (SVM). The ANFIS was used to predict how many hours the battery has been used based on two inputs i.e. load (Watt), and discharge volt (Volt). Five different prediction targets i.e. 1, 2, 3, 4, and 5 hours are selected for the ANFIS prediction. This prediction target is according to the deterioration of the discharge voltage during the measurement. The rate of voltage drop varies under different load conditions, with specific discharge profiles observed for each tested load. The result shows that ANFIS can predict the target hour based on the present load and voltage data input during the discharge operation. From the hour prediction, it can estimate the remaining useful life of the battery because the total duration of the battery is known initially. SVM is used as a comparison classifier to the ANFIS. Although SVM demonstrates superior classification accuracy (95% versus ANFIS's 88%), ANFIS's ability to predict time with two-digit precision enables more accurate remaining useful life estimation in EV applications, where even minutes of battery life can be critical for route planning and operational decisions.