As internet use increases dramatically, an increasing number of industries, including the nancial one, are operationalizing their services online. Due to the signi cant nancial losses they cause, nancial fraud is becoming a big issue as it spreads globally and grows in both volume and variety. Financial fraud detection systems should be used to identify dangers like unauthorized access and erratic assaults.During the last several years, this problem has been extensively addressed using machine learning and data mining approaches. These methods still need to be developed in order to cope with massive data, compute quickly, and spot new attack patterns. As a result, in this research, a deep learning-based approach based on the stacked temporal convolution network technique is provided for the identi cation of nancial fraud. This model aims to improve both the e ciency and accuracy of existing detection methods in the context of large data. In addition, a ower pollination optimization process is incorporated for feature selection, which addresses any side effects that may result from choosing the best features.An actual dataset of credit card, loan, and insurance frauds is used to assess the proposed model, and the results are contrasted with those of current deep learning methods. The experimental results show that the suggested FPO_QCNN achieves 99.95%, 99.95%, 99.95% of accuracy for credit card, insurance and Mortgage Data set.