One challenge when evaluating lighting distribution is dealing with the variable locations data, variability in illuminances in offices. To address the aforementioned issues: (i) when there is no daylight, a data measurement method with free and non-equally spaced personnel locations is proposed. (ii) the location information is fused, and the Gaussian function is used to numerically fit and compensate for the errors, so that a Gaussian mixture error prediction model is constructed to improve the fitting accuracy of the model. (iii) to build the best artificial lighting hybrid prediction model and improve its prediction capability. The least squares criterion is used to determine the Improved Differential Evolutionary Algorithm’s target fitness function. The prediction performance of the model was evaluated and it was found that the R2, MAE, and MSE of the proposed method were 0.996, 0.017 lx, and 0.013 lx, respectively. The method provided higher prediction accuracy and could more accurately predict the illumination state of the indoor light environment.