The prediction of biodiesel yield from animal fat oil transesterification over waste-derived solid hydroxy sodalite catalyst using artificial neural network (ANN) via a feedforward Levenberg-Marquardt (L-M) algorithm is considered in this study. A Multiple Input-Single Output (MISO) layer network with 19, 20, and 21 number of neurons in the hidden layer were tested, and the best neuron was used to construct the architecture of the network. The effect of operating variables on yield of biodiesel production over waste-derived solid sodalite catalyst was carried out using a batch reactor. In this study, 6g of the animal fat oil (AFO) was pre-heated to 110 oC to allow evaporation of the moisture content of the feedstock following other important operating conditions. The influence of alcohol (methanol)/animal fat oil (AFO) molar ratio, amount of catalyst (% waste-derived HSOD), reaction time, reaction temperature, agitation speed and effect of catalyst (waste-derived HSOD) particle size were considered. The process variables studied are catalyst weight percent (wt. %) that ranged from 0.5 to 3.0 %, HSOD catalyst, methanol-to-AFO molar ratio from 1:4 to 1:14, stirring speed (intensity) from 200 to 700 rpm and reaction temperature from 40 to 65 o C, respectively. The model performance evaluation reveal among other observations The rate of the transesterification reaction over the sodalite catalyst was increased by about 2 % when the particle size of the catalyst was reduced by about 28 %. The performance of the transesterification is also sensitive to change in temperature, where the highest temperature of 60 °C resulted in conversion of 78.3 % at the reaction time of 24 h. The rate of transesterification increased at the optimized conditions and yield of biodiesel reached 95.0 %. The neural network model architecture built using 20 number of neurons in the hidden layer gave the best results amongst the neurons studied. The result obtained from input variables representation technique with visual inspection method: Option 2, indicated that the R2 of the biodiesel yield for training, validation and testing were in the range of 89-97 %, 89-95 % and 78-91 %, respectively. The aforementioned results agreed with the results obtained from the transesterification experiments, hence; indicating the reliability of the model