Resource transaction method and device based on computing power of intelligent computing center

Abstract

The invention relates to the technical field of computing power infrastructure, and provides a resource transaction method and device based on computing power of an intelligent computing center, and the method comprises the steps: S1, obtaining transaction quotation data and transaction factor data of a to-be-processed resource, S2, inputting the transaction quotation data and the transaction factor data into a trend prediction model, and obtaining a trend prediction model; outputting the trend of the to-be-processed resources in the future preset time period; the trend prediction model runs on computing power resources of an intelligent computing center, the computing power resources are distributed computing power resources and comprise GPUs located on different nodes, and each node is provided with at least one GPU; s3, generating a transaction strategy for the to-be-processed resources according to the trend, and generating a transaction order based on the transaction strategy; and S4, determining a transaction mode of the to-be-processed resource according to the transaction order. According to the resource transaction method based on the computing power of the intelligent computing center, the efficiency of resource transaction analysis can be remarkably improved, and the accuracy of transaction strategy formulation is improved.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of computing power infrastructure, and in particular, to a resource trading method and device based on the computing power of an intelligent computing center. Background Art

[0002] With the development of artificial intelligence technology and computing power technology, the concept of intelligent computing center has emerged. "Intelligent computing center" refers to the use of large-scale heterogeneous computing power resources, including general computing power and intelligent computing power, mainly for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model reasoning and other scenarios) to provide the required computing power, data and algorithms. Intelligent computing center covers facilities, hardware, software, and can provide full-stack capabilities from bottom-level computing power to top-level application enablement.

[0003] In the current stock trading field, many existing strategy development solutions are mainly based on market data to build simple prediction models to identify potential trading opportunities and generate buying or selling strategies. Although these methods can capture price trends, valuation levels and market trends to a certain extent, their prediction capabilities and practical effects are subject to many limitations, such as insufficient data processing capabilities, limited model analysis depth, over-reliance on manual experience, and low degree of automation. In general, the prediction efficiency and accuracy of existing stock trading strategies are low. Summary of the invention

[0004] The embodiments of the present invention provide a resource trading method and device based on the computing power of an intelligent computing center to solve the technical problems of low prediction efficiency and poor accuracy of stock trading strategies existing in related technologies.

[0005] In order to solve the above-mentioned technical problems, the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present application provides a resource trading method based on the computing power of an intelligent computing center, the method comprising:

[0007] Step S1: Acquire transaction market data and transaction factor data of resources to be processed, wherein the resources to be processed include at least one of the following: stocks, futures, and bonds;

[0008] Step S2: inputting the transaction market data and the transaction factor data into a trend prediction model, and outputting the trend of the resource to be processed within a preset future time period; wherein the trend prediction model runs on the computing power resources of the intelligent computing center, and the computing power resources are distributed computing power resources, and the distributed computing power resources include: GPUs located on different nodes, and at least one GPU is set on each node;

[0009] Step S3: generating a trading strategy for the resource to be processed according to the trend, and generating a trading order based on the trading strategy;

[0010] Step S4: Determine the transaction mode of the resource to be processed according to the transaction order, wherein the transaction mode includes: buying, selling, and holding.

[0011] Optionally, step S1 includes at least one of the following:

[0012] Step S11: Acquire the transaction market data and the transaction factor data from the transaction system;

[0013] Step S12: Acquire the transaction market data from the transaction system, and calculate the transaction factor data based on the transaction market data and the distributed computing resources.

[0014] Optionally, the step S12 includes:

[0015] Step S121: acquiring the transaction market data from the transaction system, and splitting the transaction market data to obtain multiple groups of split transaction market data;

[0016] Step S122: Based on the GPUs on different nodes and the multiple groups of split trading market data, multiple groups of trading factor data are calculated in parallel, wherein each group of trading factor data corresponds to a group of split trading market data, and each GPU is responsible for processing a group of trading market data and a group of trading factor data;

[0017] Step S123: Merge the multiple groups of transaction factor data to obtain merged transaction factor data.

[0018] Optionally, step S2 includes:

[0019] Step S21: Based on the distributed computing resources of the intelligent computing center, the transaction market data and the transaction factor data are processed in parallel to obtain the processed transaction market data and the transaction factor data, wherein the processing includes at least one of the following: data cleaning and data denoising;

[0020] Step S22: extracting features from the processed transaction market data and the processed transaction factor data to obtain feature data, wherein the feature data is data associated with the trend of the resource to be processed within a preset future time period;

[0021] Step S23: input the characteristic data into the trend prediction model, and output the trend within the future preset time period.

[0022] Optionally, before step S23, the method further includes:

[0023] Step S2a: construct multiple sub-models of the trend prediction model to be trained, wherein each GPU of each node runs one of the sub-models;

[0024] Step S2b: Acquire training data, divide the training data by time, and obtain training data subsets under multiple time windows, wherein the training data is the trading market data of the sample resources, the trading factor data of the sample resources, and the trend of the sample resources in a future preset time period, and each of the training data subsets is used to train a sub-model;

[0025] Step S2c: on each GPU on each node, based on each of the training data subsets, respectively train a sub-model, and obtain the gradient information corresponding to each sub-model;

[0026] Step S2d: after the training is completed, the target gradient information is determined based on the gradient information corresponding to each sub-model;

[0027] Step S2e: Determine the parameters of the trend prediction model according to the target gradient information to obtain the trend prediction model.

[0028] Optionally, after step S1, the method further includes:

[0029] Step S5: Analyzing the risk index of the resource to be processed based on the distributed computing power resources, the transaction market data and the transaction factor data;

[0030] Step S6: generating an alarm message when the risk index exceeds a preset risk threshold, wherein the alarm message is used to notify a user of the current risk status of the resource to be processed.

[0031] Optionally, the trading market data includes at least one of the following: the opening price, closing price, highest price, lowest price, and trading volume of the resource to be processed; the trading factor data includes at least one of the following: the moving average, relative strength index, trading volume change rate, price-earnings ratio, and price-to-book ratio of the resource to be processed.

[0032] In a second aspect, an embodiment of the present application provides a resource trading device based on the computing power of an intelligent computing center, the device comprising:

[0033] An acquisition module, used to acquire transaction market data and transaction factor data of resources to be processed, wherein the resources to be processed include at least one of the following: stocks, futures, and bonds;

[0034] An execution module is used to input the transaction market data and the transaction factor data into a trend prediction model, and output the trend of the resource to be processed within a future preset time period; wherein the trend prediction model runs on the computing power resources of the intelligent computing center, and the computing power resources are distributed computing power resources, and the distributed computing power resources include: GPUs located on different nodes, and at least one GPU is set on each node;

[0035] generating a trading strategy for the resources to be processed according to the trend, and generating a trading order based on the trading strategy;

[0036] The transaction mode of the to-be-processed resource is determined according to the transaction order, wherein the transaction mode includes: buying, selling, and holding a position.

[0037] Optionally, the acquisition module is also used to obtain the transaction market data and the transaction factor data from a transaction system; obtain the transaction market data from a transaction system, and calculate the transaction factor data based on the transaction market data and the distributed computing resources.

[0038] Optionally, the acquisition module is further used to acquire the transaction market data from the transaction system, and split the transaction market data to obtain multiple groups of split transaction market data; based on the GPUs on different nodes and the multiple groups of split transaction market data, multiple groups of transaction factor data are calculated in parallel, wherein each group of transaction factor data corresponds to a group of split transaction market data, and each GPU is responsible for processing a group of transaction market data and a group of transaction factor data; the multiple groups of transaction factor data are merged to obtain merged transaction factor data.

[0039] Optionally, the execution module is further used to process the transaction market data and the transaction factor data in parallel based on the distributed computing resources of the intelligent computing center to obtain the processed transaction market data and the transaction factor data, wherein the processing includes at least one of the following: data cleaning and data denoising;

[0040] Performing feature extraction on the processed transaction market data and the processed transaction factor data to obtain feature data, wherein the feature data is data associated with the trend of the resource to be processed within a future preset time period;

[0041] The characteristic data is input into the trend prediction model, and the trend within the future preset time period is output.

[0042] Optionally, the execution module is further used to construct a plurality of sub-models of the trend prediction model to be trained before inputting the feature data into the trend prediction model and outputting the trend within the future preset time period, wherein one sub-model is run on each GPU of each node;

[0043] Acquire training data, divide the training data by time, and obtain training data subsets under multiple time windows, wherein the training data is trading market data of sample resources, trading factor data of the sample resources, and trends of the sample resources in a future preset time period, and each of the training data subsets is used to train a sub-model;

[0044] On each GPU on each node, based on each of the training data subsets, a sub-model is trained respectively, and gradient information corresponding to each sub-model is obtained;

[0045] After the training is completed, the target gradient information is determined based on the gradient information corresponding to each sub-model;

[0046] The parameters of the trend prediction model are determined according to the target gradient information to obtain the trend prediction model.

[0047] Optionally, the execution module is further configured to analyze the risk index of the resource to be processed based on the distributed computing resources, the transaction market data and the transaction factor data after acquiring the transaction market data and the transaction factor data of the resource to be processed;

[0048] When the risk index exceeds a preset risk threshold, an alarm message is generated, wherein the alarm message is used to notify a user of the current risk situation of the resource to be processed.

[0049] Optionally, the trading market data includes at least one of the following: the opening price, closing price, highest price, lowest price, and trading volume of the resource to be processed; the trading factor data includes at least one of the following: the moving average, relative strength index, trading volume change rate, price-earnings ratio, and price-to-book ratio of the resource to be processed.

[0050] In a third aspect, an embodiment of the present invention provides an electronic device comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the steps of a resource trading method based on the computing power of an intelligent computing center as described in the first aspect are implemented.

[0051] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of a resource trading method based on the computing power of an intelligent computing center as described in the first aspect are implemented.

[0052] In a fifth aspect, an embodiment of the present invention provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of a resource trading method based on the computing power of an intelligent computing center as described in the first aspect.

[0053] In the embodiments of the present invention, the distributed computing resources of the intelligent computing center are combined to provide a new direction for the development of stock trading strategies. By utilizing the powerful computing power and intelligent algorithms of the intelligent computing center, the data processing rate, model complexity and prediction accuracy can be significantly improved, thereby solving the various defects existing in the existing solutions, which not only helps to improve trading efficiency, but also helps to achieve more effective risk management and decision support in a dynamic market environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0055] Figure 1 A flowchart of a resource trading method based on the computing power of an intelligent computing center provided by an embodiment of the present invention;

[0056] Figure 2 A structural block diagram of a stock trading method provided by an embodiment of the present invention;

[0057] Figure 3 A structural block diagram of a stock trading method provided in an embodiment of the present application;

[0058] Figure 4 A flowchart of a resource trading method based on the computing power of an intelligent computing center provided by an embodiment of the present invention;

[0059] Figure 5 A flowchart of a resource trading method based on the computing power of an intelligent computing center provided by an embodiment of the present invention;

[0060] Figure 6 A flowchart of a resource trading method based on the computing power of an intelligent computing center provided by an embodiment of the present invention;

[0061] Figure 7A structural block diagram of a resource trading device based on the computing power of an intelligent computing center provided by an embodiment of the present invention;

[0062] Figure 8 FIG. 4 is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0064] The following is a brief description of the technical terms involved in the present invention.

[0065] The "computing power" mentioned in the present invention is the ability of computer equipment or computing / data center to process information. It is the ability of computer hardware and software to work together to execute certain computing requirements. It is the computing power to achieve target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity, and data storage capacity. It mainly provides services to the society through computing power infrastructure.

[0066] The "computing power (CP)" mentioned in the present invention is the ability of a data center server to process data and output results. It is a comprehensive indicator to measure the computing power of a data center, including general computing power, super computing power and intelligent computing power. The commonly used unit of measurement is the number of floating point operations per second (Floating Point Operations Per Second, FLOPS, where 1EFLOPS=10^18FLOPS). The larger the value, the stronger the comprehensive computing power. According to calculations, 1EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs (Central Processing Unit) or 2 million mainstream notebooks. The calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 In the embodiment of the present invention, the computing power adopts the half-precision floating point computing power (FP16) of the graphics card.

[0067] The "carrying capacity" (Network Power, NP) described in the present invention is a manifestation of the data transmission capability of computing facilities, including comprehensive capabilities such as network architecture, network bandwidth, transmission latency, intelligent management and scheduling, etc. Carrying capacity involves network transmission within and between data centers, and is a comprehensive indicator for measuring network transmission scheduling capabilities. In the embodiment of the present invention, the carrying capacity uses the video memory bandwidth. In the embodiment of the present invention, the storage capacity uses the video memory bandwidth.

[0068] The "Storage Power" (SP) described in the present invention refers to the comprehensive capabilities of a data center in terms of data storage capacity, performance, safety and reliability, and green and low-carbon. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and server built-in storage devices. The commonly used unit of measurement for storage capacity is exabyte (EB, 1EB = 2^60bytes), and the commonly used unit of measurement for performance is the number of reads and writes per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB). The disaster recovery ratio is an important manifestation of safety and reliability. In the embodiment of the present invention, the storage power adopts the video memory capacity.

[0069] The "computing power infrastructure" described in the present invention is a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage capacity. It can realize centralized computing, storage, transmission and application of information, and presents characteristics such as diversity and ubiquity, intelligence and agility, safety and reliability, green and low-carbon. It is of great significance to promote industrial transformation and upgrading, enable my country's scientific and technological innovation, meet people's better life and realize high-efficiency social governance.

[0070] The “computing power” mentioned in the present invention includes general computing power CP 通用 、Intelligent computing power CP 智能 And super computing power CP 超级 .

[0071] The "general computing power" mentioned in the present invention refers to the computing power provided by servers based on CPU chips, which is used to support basic general computing such as cloud computing and edge computing.

[0072] The "intelligent computing power" mentioned in the present invention refers to the large-scale deployment of intelligent computing centers based on special chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), ASIC (Application Specific Integrated Circuit) for various innovative artificial intelligence applications, such as natural language processing, machine vision, etc.

[0073] The "super computing power" mentioned in the present invention is mainly the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer intelligent computing centers working in parallel, and processes extremely complex or data-intensive problems through dedicated operating intelligent computing centers. It is mainly used for calculations in cutting-edge scientific fields, such as planetary simulation, drug molecule design, gene analysis, etc.

[0074] The "intelligent computing center" mentioned in the present invention refers to a facility that provides the required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model reasoning) by using large-scale heterogeneous computing resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from bottom-level computing power to top-level application enablement.

[0075] The "computing resources" mentioned in the present invention refer to technologies and facilities with information computing, transmission, storage and application capabilities required for the development of a digital society, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection intelligent computing centers, and supporting and guarantee resources such as wind, fire, water and electricity.

[0076] Please refer to Figure 1 , Figure 1 A resource trading method based on the computing power of an intelligent computing center provided according to an embodiment of the present invention is shown, and the method includes:

[0077] Step S1: Obtaining transaction market data and transaction factor data of resources to be processed;

[0078] The resources to be processed include at least one of the following: stocks, futures, bonds;

[0079] Step S2: Input the transaction market data and transaction factor data into the trend prediction model, and output the trend of the resource to be processed within a preset future time period;

[0080] Step S3: Generate a trading strategy for the resources to be processed according to the trend, and generate a trading order based on the trading strategy;

[0081] Step S4: Determine the transaction method of the resources to be processed according to the transaction order;

[0082] Among them, trading methods include: buying, selling, and holding positions.

[0083] And you can refer to Figure 2 and Figure 3 A structural block diagram of the stock trading method is shown.

[0084] In step S1, the transaction market data includes at least one of the following: the opening price, closing price, highest price, lowest price, and trading volume of the resource to be processed; the transaction factor data includes at least one of the following: the moving average, relative strength index, trading volume change rate, price-earnings ratio, and price-to-book ratio of the resource to be processed.

[0085] The intelligent computing center needs to collect key data for transaction analysis, mainly including: transaction market data, such as real-time market prices, trading volumes, buy and sell orders, etc., which can reflect the dynamic changes of the market. These data can come from financial products such as stocks, futures and bonds, and can be obtained in real time through high-frequency trading data streams; transaction factor data, such as various factors that affect transaction decisions, such as financial statement data, macroeconomic indicators, industry trends, technical indicators (such as moving averages, relative strength indexes, etc.). After obtaining these data, they can provide a basis for subsequent analysis and model training to ensure the accuracy and timeliness of the data.

[0086] In one possible implementation, step S1 includes at least one of the following: step S11: obtaining transaction market data and transaction factor data from the transaction system; step S12: obtaining transaction market data from the transaction system, and calculating transaction factor data based on the transaction market data and distributed computing resources.

[0087] In step S11, the intelligent computing center directly obtains dynamic market data from the trading system (exchange or trading platform). Specifically, it can be connected to the exchange through the application programming interface to obtain the latest market prices, trading volumes, buying and selling depths and other information of stocks, futures and bonds. These real-time data are crucial for market trend analysis. In addition to real-time data, the intelligent computing center (or self-built system, software platform) can also obtain historical prices and transaction data to establish a sufficient trading signal foundation to support subsequent machine learning model training. As for the extraction of trading factor data, some trading factor data (such as price-earnings ratio, financial statement data, etc.) can be directly provided by the trading platform, and the intelligent computing center extracts relevant information from it to form a trading factor library; additional industry and macroeconomic data can also be obtained from third-party providers. These data can enrich the factor library and help to conduct more comprehensive market analysis.

[0088] In step S12, transaction market data can be obtained from the trading system, and transaction factor data can be calculated based on the transaction market data and distributed computing resources. Specifically, the intelligent computing center can continuously obtain real-time market data to ensure that the latest information is used for subsequent factor calculations. Real-time market data may include monitoring price fluctuations, changes in trading volume, etc., and then define the factors that need to be calculated, such as relative strength index, moving average, capital flow index, etc. According to actual needs, set the calculation logic and parameters, and use the distributed computing resources of the intelligent computing center to process market data in parallel through multiple GPUs to achieve efficient factor calculation. Each node can independently run some computing tasks, significantly speeding up data processing.

[0089] In one possible implementation, Figure 4 As shown, step S12 includes:

[0090] Step S121: Acquire transaction market data from the transaction system, and split the transaction market data to obtain multiple groups of split transaction market data;

[0091] Step S122: Based on the GPUs on different nodes and the multiple groups of split trading market data, multiple groups of trading factor data are calculated in parallel;

[0092] Each set of transaction factor data corresponds to a set of split transaction market data, and each GPU is responsible for processing a set of transaction market data and a set of transaction factor data;

[0093] Step S123: Merge multiple groups of transaction factor data to obtain merged transaction factor data.

[0094] Therefore, by splitting the market data and relying on the computing power resources of the intelligent computing center for parallel computing, the computing speed is significantly improved. Even in a high-frequency trading environment, the system can still generate the required factor data in real time to ensure the timeliness of decision-making; by splitting the data into multiple groups, the system can flexibly adapt to different market data requests, and at the same time, it can increase and deploy computing nodes to expand computing capabilities to meet higher computing needs; after the splitting and merging process, the trading factor data is more complete and systematic, providing a reliable data basis for subsequent analysis and reducing the probability of erroneous decisions caused by data problems; the merged trading factor data will be able to quickly feed back to the trading decision-making system, supporting rapid dynamic risk control and immediate strategy adjustments, allowing the system to respond to market changes more flexibly.

[0095] In step S2, the trend prediction model runs on the computing power resources of the intelligent computing center, which are distributed computing power resources. The distributed computing power resources include: GPUs located on different nodes, and at least one GPU is set on each node. Among them, GPU is an electronic circuit specially used for processing graphics and images. Initially, GPU was designed mainly to accelerate graphics rendering, especially in video games and 3D graphics applications. However, with the development of technology, the application scope of GPU has expanded to many other fields, such as scientific computing, machine learning, and deep learning. GPU has a large number of cores and can handle multiple tasks at the same time, which makes it more efficient than CPU when processing parallel calculations.

[0096] In step S2, the model can be combined with machine learning or deep learning algorithms to analyze the input historical market data and factor data to predict future trends within a preset time period (such as the price change trend in the next hour or day). The trend prediction model runs in a distributed computing environment and can rely on GPU resources on multiple nodes for efficient computing. GPU has powerful parallel processing capabilities, can quickly process large-scale data sets and conduct complex model training, and improve the accuracy and response time of prediction.

[0097] Distributed computing resources refer to a resource allocation method that distributes computing power across multiple physical or virtual machines and connects them through a network for collaborative computing. Unlike the traditional centralized computing model, distributed computing resources can utilize the processing power of multiple computers to form a more powerful and flexible computing platform.

[0098] In one possible implementation, Figure 5 As shown, step S2 includes:

[0099] Step S21: Based on the distributed computing resources of the intelligent computing center, the transaction market data and the transaction factor data are processed in parallel to obtain the processed transaction market data and the transaction factor data;

[0100] Wherein, the processing includes at least one of the following: data cleaning, data denoising;

[0101] Step S22: extracting features from the processed transaction market data and the processed transaction factor data to obtain feature data;

[0102] The characteristic data is data associated with the trend of the resource to be processed within a preset future time period;

[0103] Step S23: input the characteristic data into the trend prediction model, and output the trend within a preset time period in the future.

[0104] It should be noted that through the distributed computing architecture of the intelligent computing center, data processing tasks can be distributed to multiple computing nodes (such as GPUs) to achieve efficient parallel processing. This method can significantly reduce data processing time and adapt to high-frequency and large-scale data requirements.

[0105] Specifically, data cleaning can be performed first to remove duplicate data. During the data collection process, duplicate information may appear, so it needs to be identified and removed. For incomplete data, appropriate methods (such as mean filling, interpolation, etc.) will be used to process missing values ​​to avoid affecting the accuracy of subsequent analysis. Statistical methods (such as standard deviation, etc.) are used to identify outliers in the data, and these outliers are removed or corrected according to specific circumstances. Then, data denoising is performed. In market conditions, data is often interfered by noise, and meaningful signals need to be identified and separated. Moving average, exponential smoothing and other methods can be used to denoise the data to improve the smoothness of the data, improve the data quality, and make subsequent analysis smoother.

[0106] Afterwards, feature extraction can be performed on the processed data to generate key variables related to future market trends. Feature extraction is the basis for the model to effectively learn and predict. Selecting appropriate features directly affects the performance and accuracy of the model. In the method shown in the embodiment of the present application, the extracted features should be able to reflect information such as price fluctuations, changes in trading volume and market trends.

[0107] Specifically, the features can be divided into the following points:

[0108] 1. Features based on technical indicators: According to the processed trading market data, the corresponding technical indicators are extracted as features, such as moving average, relative strength index, Bollinger band, etc.

[0109] 2. Extraction based on statistical features: Statistical features such as mean, standard deviation, volatility, etc. can be extracted from price data.

[0110] 3. Time series features: Generate price changes and volume changes over the past n (n is a positive integer) periods through a time window to capture potential positive trend features. These features will be used as input for machine learning models to help the model understand the relationship between past behaviors and future goals.

[0111] All extracted features should be structured by timestamp or other identifiers to facilitate integration and analysis in subsequent steps.

[0112] Then, the feature data can be input into the trend prediction model to output the trend in the future preset time period. The features extracted in the previous step can be input into the machine learning or deep learning model in the correct format to provide training and prediction support for the model. Model training is based on historical data and related features, and the model is trained using a training set. The model will learn the relationship between features and market trends. The model prediction process is to use the trained model to make predictions and generate trend assessment probabilities or price change predictions for the future preset time period. The model may output multiple prediction results at the same time to assist decision-making. The final prediction results need to be evaluated, such as using regression analysis, accuracy and other scoring criteria to verify the effectiveness of the model. Based on the evaluation results, the model is tuned and retrained to update the model parameters and improve its prediction accuracy.

[0113] In summary, the transaction data after data cleaning and denoising is more accurate, reducing potential decision-making errors caused by data problems and improving reliability; through reasonable feature extraction, complex raw data is converted into highly relevant features, which not only simplifies the data set, but also improves the model's ability to predict trends; inputting the extracted feature data into the trend prediction model enables the model to better understand market dynamics, optimize the prediction of market behavior, and improve the accuracy of investment decisions; the application of distributed computing resources makes the entire data processing process efficient and rapid.

[0114] In one possible implementation, Figure 6 As shown, before step S23, the method further includes:

[0115] Step S2a: construct multiple sub-models of the trend prediction model to be trained;

[0116] Among them, one sub-model is run on each GPU of each node;

[0117] Step S2b: Obtain training data, divide the training data by time, and obtain training data subsets under multiple time windows;

[0118] The training data includes the trading market data of the sample resources, the trading factor data of the sample resources, and the trend of the sample resources in a future preset time period. Each subset of the training data is used to train a sub-model.

[0119] Step S2c: On each GPU on each node, a sub-model is trained based on each training data subset, and the gradient information corresponding to each sub-model is obtained;

[0120] Step S2d: after the training is completed, the target gradient information is determined based on the gradient information corresponding to each sub-model;

[0121] Step S2e: Determine the parameters of the trend prediction model according to the target gradient information to obtain the trend prediction model.

[0122] In step S2a, by constructing multiple sub-models, a large amount of training data can be processed in parallel, and the prediction and generalization capabilities of the model can be improved. Different sub-models can learn different features in the data to improve the overall effect. Each sub-model can adopt a different architecture or algorithm, such as convolutional neural network, recurrent neural network, etc., and select according to specific task requirements. Different loss functions and optimization goals are set according to data characteristics to ensure that each sub-model can effectively capture task-related features. Each sub-model runs on different computing nodes, which can greatly increase the speed of model training and process larger data sets through distributed computing.

[0123] In step S2b, training data can be obtained, and the training data can be divided by time to obtain training data subsets under multiple time windows. Among them, the training data is the trading market data of the sample resources, the trading factor data of the sample resources, and the trend of the sample resources in the future preset time period. Each training data subset is used to train a sub-model. In this step, the system organizes and manages the training data so that the subsequent training process is more efficient. Specifically, the trading market data, trading factor data of the sample resources, and sample data corresponding to the future trend can be collected to synthesize the training data set, which will provide the necessary information basis for model training; then, based on market rules, the training data can be divided into fixed time periods (such as daily and weekly). For example, the data in the past year can be divided by month. The data set in each time window will form a training data subset, which can be used to train the corresponding sub-model to ensure that the model can learn the consistency and correlation of the time series. And the training data subset should ensure representativeness to avoid the degradation of model performance due to sampling bias. Comprehensively consider various market conditions and trend changes to enhance the adaptability of the sub-model.

[0124] In step S2c, each sub-model is trained using its corresponding training data subset, and the gradient of the model is calculated by the back propagation algorithm to optimize the model parameters. Specifically, each sub-model records its corresponding gradient information during the training process, and this gradient information will be used for subsequent updates of the model parameters to ensure that each sub-model can effectively converge to the optimal prediction state.

[0125] In steps S2d to S2e, the gradient information of each sub-model can be aggregated to form the final target gradient information. The target gradient information will provide guidance for the parameter update of the subsequent model, ensuring that the final model captures the knowledge learned during the multi-model training process.

[0126] In summary, different sub-models can capture different features in the data, improve the generalization and prediction capabilities of the model through diverse learning paths, and enhance the understanding of complex market behaviors; by aggregating the gradient information of each sub-model, the accuracy of model parameter updates is ensured, and the prediction uncertainty caused by the deviation of a single model is reduced; the training data is systematically divided and managed so that each sub-model can be trained on the basis of appropriate data, thereby improving the effectiveness of model learning; after multiple rounds of training and target gradient updates, the trend prediction model finally obtained has high accuracy, providing strong support for actual trading decisions; relying on the parallel training of the distributed computing resources of the intelligent computing center, different sub-models can learn independently and quickly, which improves the overall efficiency. Compared with the traditional single model training method, it greatly reduces the calculation time and improves the calculation efficiency.

[0127] In step S3 and step S4, specific trading strategies can be formulated based on the future trends output by the model using algorithms, trading orders can be determined based on the trading strategies, and then the trading methods of the resources to be processed can be determined based on the trading orders. These strategies will take into account factors such as market sentiment, trend strength, and price fluctuations. After the strategy is formulated, the intelligent computing center automatically generates corresponding trading orders (such as buy or sell), reducing the need for manual intervention and improving the automation level of transactions. The intelligent computing center will select the optimal trading method based on market conditions and strategy recommendations, and execute it at the right time to maximize profits and reduce risks.

[0128] In a possible implementation, after step S1, the method further includes:

[0129] Step S5: Analyze the risk index of the resources to be processed based on the distributed computing resources, transaction market data and transaction factor data;

[0130] Step S6: When the risk index exceeds a preset risk threshold, an alarm message is generated, wherein the alarm message is used to notify the user of the current risk status of the resource to be processed.

[0131] Therefore, combined with distributed computing resources, risk prediction models can be pre-trained, which significantly improves the speed and accuracy of risk analysis.

[0132] In summary, in the method provided by the embodiment of the present application, the GPU of the intelligent computing center has powerful parallel computing capabilities and can process thousands of data points at the same time. When applied to the training of deep learning and machine learning models, the GPU can greatly accelerate the calculation process. This significantly improves the data processing speed and computing efficiency when building complex prediction models. In stock trading, data updates frequently and real-time performance is crucial. Using GPU, the intelligent computing center can process a large amount of market data in milliseconds and quickly execute complex algorithms and models, including the calculation of risk indexes and the training of prediction models; the intelligent computing center can dynamically allocate computing resources according to actual needs. When the market fluctuates greatly, GPU resources can be increased to process higher-frequency data streams, while in stable periods, computing resources can be appropriately reduced. This flexibility effectively reduces costs and maximizes resource utilization; the training tasks of multiple models are assigned to different GPUs for parallel processing, which not only speeds up the calculation speed, but also avoids bottlenecks caused by overloading of a certain node; using distributed GPUs for model training can shorten the training time, so that investment strategies can be iterated and updated more frequently. High-frequency training and adjustment enable the model to adapt to market changes in a timely manner and effectively capture short-term market trends.

[0133] In summary, combining the distributed computing resources of the intelligent computing center provides a new direction for the development of resource (such as stocks) trading strategies. By utilizing the powerful computing power and intelligent algorithms of the intelligent computing center, the data processing rate, model complexity and prediction accuracy can be significantly improved, thereby solving the various defects in existing solutions. It not only helps to improve trading efficiency, but also helps to achieve more effective risk management and decision support in a dynamic market environment.

[0134] Figure 7 A resource trading device based on the computing power of an intelligent computing center provided according to an embodiment of the present application is shown. Figure 7 As shown, the device 70 includes:

[0135] The acquisition module 701 is used to acquire the transaction market data and transaction factor data of the resources to be processed, wherein the resources to be processed include at least one of the following: stocks, futures, and bonds;

[0136] The execution module 702 is used to input the transaction market data and the transaction factor data into the trend prediction model, and output the trend of the resource to be processed within a preset time period in the future; wherein the trend prediction model runs on the computing power resources of the intelligent computing center, and the computing power resources are distributed computing power resources, and the distributed computing power resources include: GPUs located on different nodes, and at least one GPU is set on each node;

[0137] Generate trading strategies for pending resources based on trends, and generate trading orders based on trading strategies;

[0138] The transaction method of the resources to be processed is determined according to the transaction order, where the transaction methods include: buying, selling, and holding.

[0139] In a possible implementation, the acquisition module 701 is also used to obtain transaction market data and transaction factor data from the transaction system; obtain transaction market data from the transaction system, and calculate the transaction factor data based on the transaction market data and distributed computing resources.

[0140] In a possible implementation, the acquisition module 701 is further used to obtain trading market data from the trading system, and split the trading market data to obtain multiple groups of split trading market data; based on the GPUs on different nodes and the multiple groups of split trading market data, multiple groups of trading factor data are calculated in parallel, wherein each group of trading factor data corresponds to a group of split trading market data, and each GPU is responsible for processing a group of trading market data and a group of trading factor data; and the multiple groups of trading factor data are merged to obtain merged trading factor data.

[0141] In a possible implementation, the execution module 702 is further used to process the transaction market data and the transaction factor data in parallel based on the distributed computing resources of the intelligent computing center to obtain the processed transaction market data and the transaction factor data, wherein the processing includes at least one of the following: data cleaning and data denoising;

[0142] Extracting features from the processed transaction market data and the processed transaction factor data to obtain feature data, wherein the feature data is data associated with the trend of the resource to be processed within a preset future time period;

[0143] Input the characteristic data into the trend prediction model and output the trend within a preset time period in the future.

[0144] In a possible implementation, the execution module 702 is further used to construct multiple sub-models of the trend prediction model to be trained before inputting the feature data into the trend prediction model and outputting the trend within a future preset time period, wherein one sub-model is run on each GPU of each node;

[0145] Acquire training data, divide the training data by time, and obtain training data subsets under multiple time windows, wherein the training data is the trading market data of the sample resources, the trading factor data of the sample resources, and the trend of the sample resources in a future preset time period, and each training data subset is used to train a sub-model;

[0146] On each GPU on each node, a sub-model is trained based on each training data subset, and the gradient information corresponding to each sub-model is obtained;

[0147] After the training is completed, the target gradient information is determined based on the gradient information corresponding to each sub-model;

[0148] The parameters of the trend prediction model are determined according to the target gradient information to obtain the trend prediction model.

[0149] In a possible implementation, the execution module 702 is further configured to analyze the risk index of the resource to be processed based on the distributed computing resources, the transaction market data and the transaction factor data after acquiring the transaction market data and the transaction factor data of the resource to be processed;

[0150] When the risk index exceeds a preset risk threshold, an alarm message is generated, wherein the alarm message is used to notify the user of the current risk situation of the resource to be processed.

[0151] In one possible implementation, the trading market data includes at least one of the following: the opening price, closing price, highest price, lowest price, and trading volume of the resources to be processed; the trading factor data includes at least one of the following: the moving average, relative strength index, trading volume change rate, price-earnings ratio, and price-to-book ratio of the resources to be processed.

[0152] In summary, combining the distributed computing resources of the intelligent computing center provides a new direction for the development of resource (such as stocks) trading strategies. By utilizing the powerful computing power and intelligent algorithms of the intelligent computing center, the data processing rate, model complexity and prediction accuracy can be significantly improved, thereby solving the various defects in existing solutions. It not only helps to improve trading efficiency, but also helps to achieve more effective risk management and decision support in a dynamic market environment.

[0153] Please refer to Figure 8 An embodiment of the present invention further provides an electronic device 80, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the computer program is executed by the processor 801, each process of the above-mentioned resource trading method embodiment based on the computing power of an intelligent computing center is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0154] The embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned resource trading method embodiment based on the computing power of an intelligent computing center is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0155] An embodiment of the present invention also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the various processes of the above-mentioned resource trading method embodiment based on the computing power of an intelligent computing center are implemented, and the same technical effect can be achieved. To avoid repetition, they will not be repeated here.

[0156] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0157] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0158] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A resource trading method based on the computing power of an intelligent computing center, characterized in that: The method comprises: Step S1: Acquire transaction market data and transaction factor data of resources to be processed, wherein the resources to be processed include at least one of the following: stocks, futures, and bonds; Step S2: inputting the transaction market data and the transaction factor data into a trend prediction model, and outputting the trend of the resource to be processed within a preset future time period; wherein the trend prediction model runs on the computing power resources of the intelligent computing center, and the computing power resources are distributed computing power resources, and the distributed computing power resources include: graphics processing units GPU located on different nodes, and at least one GPU is set on each node; Step S3: generating a trading strategy for the resource to be processed according to the trend, and generating a trading order based on the trading strategy; Step S4: Determine the transaction mode of the resource to be processed according to the transaction order, wherein the transaction mode includes: buying, selling, and holding.

2. The method according to claim 1, characterized in that The step S1 includes at least one of the following: Step S11: Acquire the transaction market data and the transaction factor data from the transaction system; Step S12: Acquire the transaction market data from the transaction system, and calculate the transaction factor data based on the transaction market data and the distributed computing resources.

3. The method according to claim 2, characterized in that The step S12 comprises: Step S121: acquiring the transaction market data from the transaction system, and splitting the transaction market data to obtain multiple groups of split transaction market data; Step S122: Based on the GPUs on different nodes and the multiple groups of split trading market data, multiple groups of trading factor data are calculated in parallel, wherein each group of trading factor data corresponds to a group of split trading market data, and each GPU is responsible for processing a group of trading market data and a group of trading factor data; Step S123: Merge the multiple groups of transaction factor data to obtain merged transaction factor data.

4. The method according to claim 1, characterized in that The step S2 comprises: Step S21: Based on the distributed computing resources of the intelligent computing center, the transaction market data and the transaction factor data are processed in parallel to obtain the processed transaction market data and the transaction factor data, wherein the processing includes at least one of the following: data cleaning and data denoising; Step S22: extracting features from the processed transaction market data and the processed transaction factor data to obtain feature data, wherein the feature data is data associated with the trend of the resource to be processed within a preset future time period; Step S23: input the characteristic data into the trend prediction model, and output the trend within the future preset time period.

5. The method according to claim 4, characterized in that Before step S23, the method further includes: Step S2a: construct multiple sub-models of the trend prediction model to be trained, wherein each GPU of each node runs one of the sub-models; Step S2b: Acquire training data, divide the training data by time, and obtain training data subsets under multiple time windows, wherein the training data is the trading market data of the sample resources, the trading factor data of the sample resources, and the trend of the sample resources in a future preset time period, and each of the training data subsets is used to train a sub-model; Step S2c: on each GPU on each node, based on each of the training data subsets, respectively train a sub-model, and obtain the gradient information corresponding to each sub-model; Step S2d: after the training is completed, the target gradient information is determined based on the gradient information corresponding to each sub-model; Step S2e: Determine the parameters of the trend prediction model according to the target gradient information to obtain the trend prediction model.

6. The method according to claim 1, characterized in that After step S1, the method further includes: Step S5: Analyzing the risk index of the resource to be processed based on the distributed computing power resources, the transaction market data and the transaction factor data; Step S6: generating an alarm message when the risk index exceeds a preset risk threshold, wherein the alarm message is used to notify a user of the current risk status of the resource to be processed.

7. The method according to any one of claims 1 to 6, characterized in that The transaction market data includes at least one of the following: the opening price, closing price, highest price, lowest price, and transaction volume of the resource to be processed; The transaction factor data includes at least one of the following: a moving average, a relative strength index, a trading volume change rate, a price-earnings ratio, and a price-to-book ratio of the resource to be processed.

8. A resource trading device based on the computing power of an intelligent computing center, characterized in that: The device comprises: An acquisition module, used to acquire transaction market data and transaction factor data of resources to be processed, wherein the resources to be processed include at least one of the following: stocks, futures, and bonds; An execution module is used to input the transaction market data and the transaction factor data into a trend prediction model, and output the trend of the resource to be processed within a future preset time period; wherein the trend prediction model runs on the computing power resources of the intelligent computing center, and the computing power resources are distributed computing power resources, and the distributed computing power resources include: GPUs located on different nodes, and at least one GPU is set on each node; generating a trading strategy for the resources to be processed according to the trend, and generating a trading order based on the trading strategy; The transaction mode of the to-be-processed resource is determined according to the transaction order, wherein the transaction mode includes: buying, selling, and holding a position.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of a resource trading method based on the computing power of an intelligent computing center as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a resource trading method based on the computing power of an intelligent computing center as described in any one of claims 1 to 7.

11. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of a resource trading method based on the computing power of an intelligent computing center as described in any one of claims 1-7.

Images