Comparative ANFIS Models for Stochastic On-road Vehicle CO2 Emission using Grid Partitioning, Subtractive, and Fuzzy C-means Clustering

On-road vehicle CO2 emission is stochastic and is presently not feasible to be solved using hard computing methodologies due to computational cost. This paper presents an on-road paratransit vehicle CO2 emission estimation model using an adaptive neuro-fuzzy inference system (ANFIS). With input parameters, namely, the speed, slope, and acceleration, three ANFIS clustering types were utilized. Results have shown that Fuzzy-C means clustering method (FCM) obtained the best performance concerning error rates and computation simplicity. Specifically, it has yielded 13.38% NRMSE using five membership functions per input and five fuzzy rules. The grid partitioning (GP) obtained the worst prediction output while the subtractive clustering method (SCM) has comparable prediction accuracy with FCM but has a higher computational cost compared to the latter. The proposed estimation model is beneficial for paratransit vehicles wherein the state-of-the-art on-road emission models are deemed unsuitable.