To address the problem of poor localization performance caused by the small number of fingerprint points and the lack of representativeness of fingerprint features in the offline phase of fingerprint-based indoor localization, a CSI-based indoor localiza‐ tion method is proposed using Inverse Distance Weighted (IDW) interpolation and amplitude-phase fusion network. The IDW inter‐ polation algorithm is used to generate a large-capacity fingerprint library. Then, a parallel convolutional neural network (CNN) is used to process the amplitude and phase to obtain the positional fingerprint features. Finally, a novel integrated architecture fusing Random Forest (RF) and Multilayer Perceptron (MLP) is used for classification to obtain the estimated positions of the target loca‐ tion samples.