In this paper, a neuro-fuzzy-based method for analysing odour generation factors to the perception of livestock farm odour was proposed. The proposed approach incorporates neuro-adaptive learning techniques into fuzzy logic method. Rather than choosing the parameters associated with a given membership function by trail and error, these parameters could be tuned automatically in a systematic manner so as to adjust the membership functions of the input/output variables for optimal system performance. A multi-factor livestock farm odour model was developed, and both numeric factors and linguistic factors were considered. The proposed approach was tested with a livestock farm odour database.  It can incorporate non-numeric data and subjective human expert knowledge, which allows prior knowledge to be included in the model. In addition, the membership functions can be tuned during the learning according to input/output data to optimise performance. This feature is important as it can avoid inappropriate predetermination of membership functions, thus minimising errors resulting from the limited knowledge of the livestock farm odour system. A livestock farm database collected by our research team has been used in this study. The results show that the proposed approach is effective and provides a much more accurate odour prediction in comparison with a typical multi-layer feedforward neural network.