The Bacillus cereus group comprises genetical closely related species with variable toxigenic characteristics. However, detection and differentiation of the B. cereus group species in routine diagnostics can be difficult, expensive and laborious since current species designation is linked to specific phenotypic characteristic or the presence of species-specific genes. Especially the differentiation of Bacillus cereus and Bacillus thuringiensis , the identification of psychrotolerant Bacillus mycoides and Bacillus weihenstephanensis , as well as the identification of emetic B. cereus and Bacillus cytotoxicu s, which are both producing highly potent toxins, is of high importance in food microbiology. Thus, we investigated the use of a machine learning approach, based on artificial neural network (ANN) assisted Fourier transform infrared (FTIR) spectroscopy, for discrimination of B. cereus group members. The deep learning tool box of Matlab was employed to construct a one-level ANN, allowing the discrimination of the aforementioned B. cereus group members. This model resulted in 100% correct identification for the training set and 99.5% correct identification overall. The established ANN was applied to investigate the composition of B. cereus group members in soil, as a natural habitat of B. cereus , and in food samples originating from foodborne outbreaks. These analyses revealed a high complexity of B. cereus group populations, not only in soil samples but also in the samples from the foodborne outbreaks, highlighting the importance of taking multiple isolates from samples implicated in food poisonings. Notable, in contrast to the soil samples, no bacteria belonging to the psychrotolerant B. cereus group members were detected in the food samples linked to foodborne outbreaks, while the overall abundancy of B. thuringiensis did not significantly differ between the sample categories. None of the isolates was classified as B. cytotoxicus , fostering the hypothesis that the latter species is linked to very specific ecological niches. Overall, our work shows that machine learning assisted (FTIR) spectroscopy is suitable for identification of B. cereus group members in routine diagnostics and outbreak investigations. In addition, it is a promising tool to explore the natural habitats of B. cereus group, such as soil.