Machine learning methods have been widely used to support decision-making, but most of the time, decisions cannot be easily explained. Therefore, providing explanations about the results generated by them becomes important. This is particularly relevant in high-risk decision scenarios in order to protect all the participants, as occurs, for example, in financial applications such as credit analysis. This work proposes CausalBioCF, a new method of explainability of machine learning algorithms based on counterfactuals. CausalBioCF combines a low-cost bioinspired optimization technique with domain-knowledge causal relationships to find more viable counterfactuals. The main novelty of CausalBioCF is the integration of counterfactual generation methods and traditional causality analysis. Compared to state-of-the-art systems such as DiCE, our experimental evaluation shows that CausalBioCF performs favorably, particularly when metrics such as sparsity and distance are considered. Furthermore, we show that existing systems such as DiCE can be improved by taking into account causal relationships.