Important global systemic banks that are extremely active in the fiscal request, similar as Nordea, have knockouts of thousands of deals settled every day. It's relatively not bring-effective to control each sale manually by the regulator, which may also bring homemade crimes( similar as different regulators giving different judgments on the same sale).
In particular, this process requires control labor force to have extremely rich trading and request experience and numerous regulators are former dealers of the bank. Arranging these workers on such a time- consuming, laborious and repetitious work isn't only a great waste of mortal coffers, but also not conducive to the bank for retaining these workers. In the bank database, there are knockouts of millions of complete records of abnormal data.
We can use these data to train and fit a deep neural network to prognosticate sale control. We eventually enforced a deep neural network in which each sale and the real- time price data related to it are treated as a sample, and the final affair of this model is a Boolean variable, that is, whether the sale is an abnormal sale.
At the same time, we integrate this model into the being business process. When a new sale is reached, the deep neural network first predicts whether the sale is abnormal, and also the results are reviewed by educated regulators to decide whether to borrow the machine’s suggestion.
All the final results will be saved in the training sample library of the network and will be included in the training sample when the model is recalibrated coming time( the being model is recalibrated every two hours
