Machine learning plays an important role in WiFi fingerprint localization techniques. To address the problem that the effect of signal fluctuation on fingerprint recognition is often ignored and how to extract broader representation information from samples, this article proposes a WiFi localization algorithm based on improved contrastive learning and parallel fusion neural network. Firstly, the algorithm utilizes the improved CL to improve fingerprint discrimination, which increases the differentiation between different categories of fingerprints while reducing the differences between fingerprints of the same category. Secondly, a parallel fusion network based on CNN and LSTM is established. Compared with the traditional serial fusion method, the network can extract more effective features from the original samples. In addition, a flatten layer is added after the pooling layer to further consider the intermediate layer information of the network. Thus, a wider range of feature information is utilized to improve the generalization performance of the model. The results show that the proposed algorithm improves the localization performance by 26% over other localization algorithms.