Method and system for high frequency trading

1. A method for high frequency trading, wherein the method is performed by one or more processors comprising a first processor for pre-processing and a second processor for a machine learning model and the first processor and the second processor are separate hardware, the method comprising:

receiving market data for a target item during a first unit time;

calculating a latency for a market order for each of a plurality of candidate batch sizes;

selecting a batch size from among the plurality of candidate batch sizes based on the calculated latency;

generating, by the first processor, input data corresponding to the selected batch size using market data for the target item;

transmitting, by the first processor, the input data to the second processor;

receiving, by the second processor, the input data;

generating, by the second processor, using the machine learning model, prediction data for the target item at a future time point, in the future time of the first unit time, associated with the selected batch size, based on the generated input data; and

generating order data for the target item based on the generated prediction data.

2. The method of claim 1 , wherein the selecting the batch size includes, for each of a plurality of predetermined future time points, selecting, from among the plurality of candidate batch sizes, a largest candidate batch size of which an end time point of the calculated latency precedes each of the plurality of predetermined future time points.

3. The method of claim 2 , wherein the selecting the batch size includes:

calculating an anticipated profit for each of selected batch sizes for each of the plurality of predetermined future time points; and

selecting a batch size having the highest calculated anticipated profit form among the selected batch sizes for each of the plurality of predetermined future time points.

4. The method of claim 3 , wherein the calculating the anticipated profit includes, for each of the plurality of predetermined future time points, calculating an anticipated profit of each of the selected batch sizes based on each of the selected batch sizes, a profit per query, and a computation time of the machine learning model for each of the selected batch sizes.

5. The method of claim 1 , wherein the calculating the latency includes, for each of the plurality of candidate batch sizes, calculating the latency based on at least one of a data rate, bandwidths of input and output data between the first processor and the second processor, sizes of the input and output data, a computation speed of the machine learning model by the second processor, a processing speed of the first processor, or a busy state of the second processor.

6. The method of claim 5 , wherein the latency includes a time taken for the market data to be pre-processed by the first processor, a time taken for transferring the pre-processed data from the first processor to the second processor, a time taken for computation of the machine learning model to be completed by the second processor, a time taken for a computation result to be transferred from the second processor to the first processor, and a time taken for the order data to be generated based on the computation result by the first processor.

7. The method of claim 1 , further comprising acquiring data precision of the input data, and

wherein the calculating the latency includes calculating, for each of the plurality of candidate batch sizes, a latency based on the precision of the input data.

8. The method of claim 7 , wherein the latency is calculated based on processing elements that are increased as the precision of the input data is decreased.

9. The method of claim 7 , wherein the selecting the batch size includes selecting a batch size from among the plurality of candidate batch sizes based on the acquired data precision of the input data.

10. The method of claim 9 , wherein the batch size is selected based on processing elements that are increased as the precision of the input data is decreased.

11. The method of claim 1 , wherein the second processor includes first and second dedicated accelerators for processing computation of the machine learning model,

the calculating the latency includes calculating, for each of the plurality of candidate batch sizes, a first latency including a computation time of the first dedicated accelerator and a second latency including a computation time of the second dedicated accelerator, and

the selecting the batch size includes selecting, for each of a plurality of predetermined future time points, a largest candidate batch size of which an end time point of a lower latency of the calculated first and second latencies precedes each of the plurality of predetermined future time points.

12. A non-transitory computer-readable recording medium storing instructions that, when executed by one or more processors, cause performance of the method according to claim 1 .

13. A high frequency trading system, comprising:

a first memory storing one or more instructions;

one or more processors comprising a first processor for pre-processing and a second processor for a machine learning model and the first processor and the second processor are separate hardware, the one or more processors configured to, by executing the one or more instructions in the first memory, receive market data for a target item during a first unit time, calculate a latency for a market order for each of a plurality of candidate batch sizes, select a batch size from among the plurality of candidate batch sizes based on the calculated latency, and generate, by the first processor, input data corresponding to the selected batch size using market data for the target item, transmit, by the first processor, the input data to the second processor;

a second memory storing one or more instructions; and

the second processor configured to, by executing the one or more instructions in the second memory, receive the input data, generate, using the machine learning model, prediction data for the target item at a future time point associated with the selected batch size based on the generated input data, and provide the generated prediction data to the one or more processors,

wherein the one or more processors are further configured to generate order data for the target item based on the prediction data provided from the second processor.

14. The high frequency trading system of claim 13 , wherein the calculating the latency includes, for each of the plurality of candidate batch sizes, calculating the latency based on at least one of a data rate, bandwidths of input and output data between the first processor and the second processor, sizes of the input and output data, a computation speed of the machine learning model by the second processor, a processing speed of the first processor, or a busy state of the second processor.

15. The high frequency trading system of claim 14 , wherein the latency includes a time taken for the market data to be pre-processed by the first processor, a time taken for transferring the pre-processed data from the first processor to the second processor, a time taken for computation of the machine learning model to be completed by the second processor, a time taken for a computation result to be transferred from the second processor to the first processor, and a time taken for the order data to be generated based on the computation result by the first processor.

16. The high frequency trading system of claim 13 , wherein the one or more processors are further configured to acquire data precision of the input data, and

the calculating the latency includes calculating, for each of the plurality of candidate batch sizes, a latency based on the precision of the input data.

17. The high frequency trading system of claim 16 , wherein the latency is calculated based on processing elements that are increased as the data precision of the input data is decreased.

18. The high frequency trading system of claim 16 , wherein the selecting the batch size includes selecting a batch size from among the plurality of candidate batch sizes based on the acquired data precision of the input data.

19. The high frequency trading system of claim 18 , wherein the batch size is selected based on processing elements that are increased as the precision of the input data is decreased.

20. The high frequency trading system of claim 13 , wherein the second processor includes first and second dedicated accelerators for processing computation of the machine learning model,

the calculating the latency includes calculating, for each of the plurality of candidate batch sizes, a first latency including a computation time of the first dedicated accelerator and a second latency including a computation time of the second dedicated accelerator, and

the selecting the batch size includes selecting, for each of a plurality of predetermined future time points, a largest candidate batch size of which an end time point of a lower latency of the calculated first and second latencies precedes each of the plurality of predetermined future time points.