<p>Spatial ontology is one of the significant techniques for detecting suspicious transactions in financial institutions. This work proposes an expert system via spatial ontology-based Deep learning for effectively detecting abnormal transactions. The proposed methodology has the stages like pre-processing, construction of ontology and transaction detection. The initial step is pre-processing, during which specific data items are converted into ontology format and noisy data is eliminated. Next, a set of semantic web rule languages and a knowledge base are incorporated into the construction of ontology. New knowledge is gathered from the predefined rules using native reasoning to assist in ontology about every transaction. The proposed method uses a stacked autoencoder (SAE), a deep learning technique, to classify normal and abnormal transactions, and it is trained on the deduced knowledge base. After the classification process, the abnormal transaction list is provided to the query inferred to obtain different information from that list. The experimental analysis is carried out on the two benchmark datasets, synthetic financial and credit card datasets, and compared with the existing approaches. The accuracy obtained by the proposed model for synthetic financial and credit card datasets is 93% and 96.23%, respectively. The experimental results showed that the proposed system performed better than the state-of-the-art approaches.</p>