This study aims to introduce a novel semi-supervised learning algorithm in fuzzy min-max neural networks, known as FMN-OE, focusing on calculating and eliminating the potential overlap between hyperboxes in the network. Overlap between hyperboxes can lead to ambiguity and uncertainty in recognition and classification, potentially reducing the performance of FMNN. To address this issue, the proposed methods involve adjusting and shrinking overlapping hyperboxes to eliminate the overlap. However, this adjustment may result in the loss of valuable information when removing data points that provide additional context from the previously evaluated hyperbox. To mitigate this drawback, FMN-OE evaluates whether the expansion adjustments will create overlapping regions between hyperboxes. If overlap is detected, FMN-OE adjusts the hyperboxes to expand them while striving to avoid creating overlapping regions. Finally, a new hyperbox is generated from the training data that causes overlap. FMN-OE determines supplementary information for this new hyperbox based on fuzziness and/or additional information provided by experts during the training process. The performance and superiority of the proposed method on synthetic and Benchmark datasets are demonstrated by comparing it with existing unsupervised and semi-supervised fuzzy clustering methods operating under similar conditions.