Introduction. The article notes that current interval forecasting methodology is often limited to only three types of models: linear, quadratic, and cubic. However, for certain datasets, these models may prove too rigid and fail to respond flexibly to underlying fluctuations. This paper demonstrates the feasibility of applying logarithmic forms of the exponential model and the logarithmic parabola in interval forecasting. The study investigates how logarithmization affects the quality of data fitting and the accuracy of forecast error estimation. Furthermore, practical recommendations are provided for implementing logarithmic models in criminological forecasting. Materials and Methods. The research draws upon Russian national crime statistics and utilizes advanced mathematical statistical methods. The Results of the Study. The study proposes expanding the toolkit for interval criminological forecasting by incorporating two additional models: the exponential model and the logarithmic parabola. These models were empirically tested for short-term forecasting of registered crime rates in the Krasnoyarsk Territory of the Russian Federation. The findings indicate that logarithmization significantly enhances the quality of data fitting and improves the predictive performance of the models. Findings and Conclusions: The study confirms the efficacy of using logarithmic models for criminological crime forecasting. When applying these models, it is essential to evaluate the goodness-of-fit of the back-transformed (original-scale) models. The selection of a model for criminological forecasting should be based on the predictive performance of the back-transformed models rather than the logarithmic ones. Forecast error estimates for these models increase significantly as the data values grow; this must be taken into account when forecasting processes with an upward trend.