In the era of rapid Internet technological advancement, the scale of online transactions is incessantly expanding. Concurrently, the issue of network transaction fraud has attained heightened significance. In contrast to credit card transactions, online transactions exhibit characteristics such as low cost, extensive coverage, and high frequency, rendering fraud detection a notably intricate challenge. This paper addresses the complexities associated with fraud detection in online transactions by proposing two distinct algorithms: one based on a Fully Connected Neural Network and the other utilizing XGBoost. These algorithms demonstrate commendable performance, with AUC values reaching 0.912 and 0.969, respectively.