Anchor point selection method and system for asset price prediction

Abstract

The invention provides an anchor point selection method and system for asset price prediction. The method comprises the following steps: receiving historical price data of target assets and candidate assets in a predetermined period; determining the correlation between the candidate assets and the target assets based on the historical price data of the target assets and the candidate assets; determining the sensitivity of the target asset to the price fluctuation of the candidate assets for the determined candidate assets with high correlation; and taking the determined candidate assets with high sensitivity as anchor assets, so as to predict the asset price of the target asset based on the anchor assets.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to prediction problems in general, and to classification prediction in particular. BACKGROUND

[0002] Finance plays a very important role in the economy. Since the development of finance, many basic theories and systems, as well as various models and tools, have been formed to study the operation rules of economy and finance from different perspectives. In finance, investment and trading are all centered on various assets. In modern finance, various theories and models for asset pricing have been developed, including the representative theories and models such as the Capital Asset Pricing Model (CAPM), which was developed by American scholars William Sharpe, John Lintnor, Jack Treynor, and Jan Mossin in 1964 based on portfolio theory and capital market theory, mainly studying the relationship between expected return and risk of assets in the securities market, and how the equilibrium price is formed. The capital pricing model, together with option pricing theory and corporate asset structure, is known as the three pillars of modern finance theory.

[0003] However, the price of an asset is actually the result of the comprehensive consideration of various factors by the market, including asset supply and demand relationship, international and domestic political situation, macroeconomic policy, natural disasters, major events, market sentiment, and so on. Therefore, the law and prediction of asset price changes have been an important research topic in economics and finance for many years.

[0004] A typical asset price prediction method is to analyze the historical trend of asset prices based on historical data of asset prices, and to predict the future trend of asset prices. Another prediction method is to study the influence of other non-price factors on asset prices, and to predict the changes of asset prices through these factors, such as the stock market prediction method based on sentiment analysis and hidden Markov model fusion described in the authorized Chinese patent (ZL201410023154.2).

[0005] In recent years, with the development of computer science, especially the rapid development of computing power, tools such as big data and artificial intelligence technology have been introduced into the financial field, and have tried to solve many difficult problems that have plagued for many years. However, due to the characteristics of finance and economics, in terms of asset price prediction, there are many factors that may affect asset prices and are complex. It is amazing to train a prediction model that takes into account all factors, and even with the current hardware capabilities, it is not realistic.

[0006] Therefore, when performing data analysis and information processing in the financial field by using a computing tool, how to more effectively use computing resources to improve analysis accuracy and save resource consumption and improve processing efficiency as much as possible is a technical problem currently faced by the financial technology field. SUMMARY

[0007] To solve the above technical problems, the present disclosure provides an anchor point selection method and system for asset price prediction and an anchor point-based asset price prediction method and system. In the face of complex and variable prediction objects, suitable anchor assets can be selected through rapid data acquisition and data processing, and the price change relationship between the prediction object and the anchor asset can be found based on the anchor asset, thereby improving the accuracy and efficiency of prediction.

[0008] In an embodiment of the present disclosure, an anchor point selection method for asset price prediction is provided, comprising: receiving historical price data of a target asset and candidate assets in a predetermined period; determining the correlation of the candidate assets with the target asset based on the historical price data of the target asset and the candidate assets; determining the sensitivity of the target asset to asset price fluctuations of the candidate assets for the candidate assets with high correlation determined; and determining the candidate assets with strong sensitivity as anchor assets for predicting the asset price of the target asset based on the anchor assets.

[0009] In another embodiment of the present disclosure, the received historical price data of the target asset and the candidate assets in a predetermined period constitutes an asset data set.

[0010] In yet another embodiment of the present disclosure, the determination of the correlation of the candidate assets with the target asset is based on a correlation data set, which is generated by processing the asset data set.

[0011] In another embodiment of the present disclosure, the determination of the sensitivity of the target asset to asset price fluctuations of the candidate assets is based on a sensitivity data set, which is generated by processing the correlation data set.

[0012] In yet another embodiment of the present disclosure, the determination of the candidate assets with strong sensitivity as anchor assets further comprises further screening among the candidate assets with strong sensitivity based on an anchor asset data set, which is generated by processing the sensitivity data set.

[0013] In another embodiment of the present disclosure, the determination of the correlation of the candidate assets with the target asset comprises classification of the candidate assets based on the correlation data set, and the classification is binary classification or multi-classification or two-level or multi-level.

[0014] In yet another embodiment of the present disclosure, determining the sensitivity of the target asset to asset price fluctuation of the candidate asset comprises further classifying the candidate asset based on the sensitivity dataset, the classification being binary or multi-class or binary level or multi-level.

[0015] In another embodiment of the present disclosure, the predetermined period is selected on-demand.

[0016] In yet another embodiment of the present disclosure, the anchor asset can be one asset or a combination of multiple assets.

[0017] In an embodiment of the present disclosure, there is provided a method for anchor selection based asset price prediction, comprising determining an anchor asset for a target asset by the method as described above; and predicting the asset price of the target asset based on the anchor asset.

[0018] In an embodiment of the present disclosure, there is provided a system for anchor selection for asset price prediction, comprising: a receiving module configured to receive historical price data of a target asset and candidate assets for a predetermined period; a correlation determining module configured to determine correlation of the candidate assets with the target asset based on the historical price data of the target asset and the candidate assets; a sensitivity determining module configured to determine sensitivity of the target asset to asset price fluctuation of the candidate assets for the candidate assets determined to have high correlation; and an anchor selection module configured to select the candidate assets determined to have strong sensitivity as anchor assets for predicting the price of the target asset based on the anchor assets.

[0019] In an embodiment of the present disclosure, there is provided a system for anchor based asset price prediction, comprising: a receiving module configured to receive historical price data of a target asset and candidate assets for a predetermined period; a correlation determining module configured to determine correlation of the candidate assets with the target asset based on the historical price data of the target asset and the candidate assets; a sensitivity determining module configured to determine sensitivity of the target asset to asset price fluctuation of the candidate assets for the candidate assets determined to have high correlation; an anchor selection module configured to select the candidate assets determined to have strong sensitivity as anchor assets; and a prediction module configured to predict the price of the target asset based on the anchor assets.

[0020] In an embodiment of the present disclosure, there is provided a computer readable storage medium having computer instructions stored thereon, which when executed implement the method as described above.

[0021] This Summary is provided to introduce some concepts in a simplified form, further described below in the DETAILED DESCRIPTION. This Summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF DRAWINGS

[0022] The above summary of the present disclosure and the following detailed description of specific embodiments of the present disclosure will be better understood when read in conjunction with the appended drawings, where like reference numerals refer to like elements in the several figures of the drawings.

[0023] Figure 1 is a schematic diagram illustrating anchor points or anchor boxes in an image processing scenario;

[0024] Figure 2 is a schematic diagram illustrating anchor point or anchor box based prediction in an image processing scenario;

[0025] Figure 3 is a flowchart illustrating an anchor point selection method for asset price prediction according to an embodiment of the present disclosure;

[0026] Figure 4 is a flowchart illustrating an anchor point based asset price prediction method according to an embodiment of the present disclosure;

[0027] Figure 5 is a schematic diagram illustrating an anchor point selection and anchor point based asset price prediction process according to an embodiment of the present disclosure;

[0028] Figure 6 is a dataflow diagram illustrating an anchor point selection and anchor point based asset price prediction process according to an embodiment of the present disclosure;

[0029] Figure 7 is a block diagram illustrating an anchor point selection system for asset price prediction according to an embodiment of the present disclosure;

[0030] Figure 8 is a block diagram illustrating an anchor point based asset price prediction system according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0031] In order to make the above objectives, features and advantages of the present disclosure more clear and easily understood, the following will describe the specific embodiments of the present disclosure in detail with reference to the accompanying drawings.

[0032] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent to one skilled in the art that the present disclosure can be practiced without the specific details and other implementations can be employed. Thus, the present disclosure is not limited to the embodiments disclosed below but only by the claims.

[0033] Prediction is a discipline that uses inertia principle, analogy principle and other related theories to predict the future development trend of some unknown attributes or known attributes of things based on existing data (including information and data, etc.). Its core is to establish a suitable prediction model to describe the relationship between unknown and known. In real life, the application of prediction is very wide, for example, predicting the future development trend of an industry, the future sales of a store, the number of people infected with a virus in the future, etc.

[0034] Image recognition is essentially a prediction applied to image processing. Image object detection is a task of "where is what", in which the class, number, location and size of the target are uncertain. Traditional non-deep learning methods and early deep learning methods all use pyramid multi-scale plus traversal sliding window to judge "whether there is a recognized target at this scale and this position" scale by scale and position by position, which is very cumbersome and time-consuming.

[0035] And the recent target detection methods almost all use anchor technology. Figure 1 is a schematic diagram showing anchor points or anchor boxes in the image processing scene. Anchor originally refers to the iron anchor for fixing the ship, while in computer vision, it refers to anchor points (such as Figure 1 left side of the figure) or anchor boxes (such as Figure 1 right side of the figure). Anchor box is a fixed reference box commonly used in target detection. Anchor technology converts the problem into "whether there is a recognized target in this fixed reference box, and how far the target box deviates from the reference box", which no longer needs to traverse the sliding window in multiple scales, and truly realizes good and fast, for example, anchor is an important part in the target detection framework of Faster R-CNN and SSD and its extended algorithms.

[0036] In target detection, the model not only needs to learn the class of the target, but also needs to learn the position, size or target feature (such as specific endogenous topological structure) of the target. Figure 2 is a schematic diagram showing prediction based on anchor points or anchor boxes in the image processing scene. As shown in Figure 2 left side of the figure, based on the position, size or target feature of the anchor point, (δx, δy) is calculated for the position, size or target feature of the target, and the prediction result is obtained.

[0037] As shown in Figure 2As shown in the right side of the figure, anchor boxes of different sizes and aspect ratios are used (as shown, three anchor boxes of different sizes and aspect ratios, and specific endogenous topological structures are data information converted from relevant asset price information factors). By setting anchor boxes of different scales, there is a higher probability of finding prior boxes that have a good matching degree for the target object (reflected in high IOU). That is, a larger intersection over union (IOU) can be obtained.

[0038] In the prediction scenario of non-image processing, similarly, anchor-based prediction is no longer data traversal for all objects, but rather a reference system with more prominent features (for example, anchor assets) is selected for better and faster prediction.

[0039] In modern finance, there are many asset pricing models, theories and asset price evaluation methods, and based on these models, theories and assets, many prediction methods for future asset prices and trends have been developed. In recent years, more and more research has found that the price and trend of an asset is more or less affected by the price and trend of one or more other assets, and the followability of different assets is much greater than the certainty of the trend of asset prices themselves. Therefore, compared to studying the rules of a single target asset based on its historical data and making predictions, predicting the price and trend of a target asset based on the followability rules of other assets and the relevance and impact of the target asset is more scientific and has higher computational efficiency.

[0040] Based on this, the present disclosure provides an anchor-based asset price prediction scheme, which can obtain anchor points through fast data acquisition and data processing when facing complex and variable prediction objects, and can find the relationship between the prediction object and the anchor point or the quantitative relationship between multiple factors, thereby improving the accuracy and efficiency of prediction.

[0041] In the present disclosure, the specific description of the scheme will be mainly described taking asset price prediction as an example. Those skilled in the art can understand that the anchor-based prediction scheme of the present disclosure is applicable to various application scenarios and is not limited to asset price prediction.

[0042] Figure 3 is a flow chart showing an anchor selection method 300 for asset price prediction according to an embodiment of the present disclosure. As shown in the figure, the anchor selection method 300 includes the following steps: Figure 3As shown in FIG. 3, the method 300 starts at step 302, where historical price data of a target asset and candidate assets within a predetermined period are received. The types of assets are numerous, and in the context of the present disclosure, an asset can refer to a capital asset that can be traded in the market, such as a security (including individual stocks and composite indices), a commodity (such as gold, crude oil), and the like. A "target asset" refers to an asset that is the object of analysis and prediction. In contrast, an "anchor asset" refers to a reference point (system) that is used to observe, measure, and influence the target asset. A "candidate asset" is an asset that is selected as a candidate for comparison and selection in the process of determining the anchor asset, and the selection of the candidate asset can be arbitrary or based on general experience and common sense. For example, in the case of considering the NASDAQ Composite Index as a target asset, according to experience and common sense, it is easy to think that the S&P 500 Index, the S&P 100 Index, and the Dow Jones Index of the United States can be suitable as anchor assets, and thus can be selected as candidate assets in order to select one or more that are most suitable as anchor assets through comparison. The number of candidate assets is not limited, and for example, in another example, in the case of considering the ChiNext Index as a target asset, in addition to the Shanghai Composite Index and the Shenzhen Component Index, the Dow Jones, S&P, and NASDAQ indices of the United States, and various indices of the securities markets of China Hong Kong, Japan, Germany, the United Kingdom, and the like can also be selected as candidate assets.

[0043] According to one embodiment of the present disclosure, the historical price data of the assets can be received from a plurality of different data sources. The data sources can be official data sources provided by various asset exchanges (such as securities exchanges, commodity exchanges), or can be data sources of third-party data service providers authorized to provide historical price data of assets. Through the data interface provided by the data source, the historical price data of the assets can be automatically or as needed downloaded from the server of the data source to the local by a computer (such as a configured data collection module).

[0044] The received historical data can be full volume data, but generally, too old data has limited reference value and consumes bandwidth and resources. Therefore, a time period can be preset, for example, the last 10 years, 20 years, the last 10 trading days, or any other time period. According to an embodiment of the present disclosure, the predetermined period can be selected on demand. For example, the setting of this period can be related to the data required in the subsequent algorithm for determining the degree of price influence between assets and predicting asset prices. In addition, the selection of the predetermined period can also include consideration of special events that have occurred in history, for example, it is advantageous to select historical data before and after a financial crisis or other major event as a research sample, because generally such major events can induce relatively obvious changes and fluctuations in asset prices, thus facilitating the observation and analysis of the correlation and influence degree between different assets. Similarly, the time period can also be cut on demand. For example, in a research period of one year, it can be cut into four equal parts according to the quarters, or it can be cut into 12 equal parts according to the months. Those skilled in the art can understand that such cutting can also be selected on demand. Based on the cutting of a specific time period, it is advantageous to improve the accuracy of prediction before and after similar events occur.

[0045] Optionally, the historical price data received from different data sources can be preprocessed, for example, the format of the data is unified, including the unification of the time, currency unit, etc. of the data, and the extraction, deletion or combination of different fields to convert into the data structure required in the subsequent processing process. Of course, preprocessing can also include coordinate transformation of data time period and standardization of data.

[0046] According to an embodiment of the present disclosure, the received historical price data of the target asset and the candidate asset within the predetermined period can constitute an asset data set. The data in the asset data set can include asset price raw data and derived data reflecting mathematical or statistical properties of asset prices based on price data processing. For example, the data items in the asset data set can include but are not limited to: asset prices, price ratios, asset price volatility, annualized mean square deviation, normal distribution, asset return rate net variance, etc. It can be understood that the data types provided by different data sources can be different, in addition to providing asset price raw data, some data sources can provide part of the derived data. Through data preprocessing processes such as the above, the received raw data and available derived data can be extracted, and the rest of the subsequent required derived data can be obtained through local processing, and finally the asset data set associated with the target asset and each candidate asset is obtained.

[0047] At step 304, the correlation between the candidate assets and the target asset is determined based on historical price data of the target asset and the candidate assets. After the candidate assets are selected, the correlation between the candidate assets and the target asset can be analyzed first, and a preliminary screening can be performed based on the correlation to screen out candidate assets with higher correlation with the target asset, which will be conducive to saving the calculation resources and time consumed in subsequent analysis and processing.

[0048] According to an embodiment of the present disclosure, the correlation between the candidate assets and the target asset can be determined based on a correlation data set. The correlation data set can include various data reflecting the correlation between assets and assets, also known as indicators, parameters or variables. The data in the correlation data set can include but is not limited to: the volatility and correlation coefficient of two assets, the integrated volatility of two assets, the option price of two assets, the option difference, the integrated mean square error of two assets, the net mean square error of two assets relative to the integrated mean square error, etc.

[0049] The data items in the correlation data set can be generated by processing the aforementioned asset data set. It can be understood that the correlation reflects various degrees of correlation between two assets, which can be quantified, so based on the historical data of the target asset and each candidate asset in the asset data set obtained by big data technology, the volatility characteristics of each asset itself and the linkage characteristics between the volatilities of different assets in the same time dimension can be analyzed by mathematical operation processing on time series data such as the historical data of two assets, and the results of mathematical analysis are used as the correlation data between each candidate asset and the target asset.

[0050] As an example, but not limitation, some exemplary acquisition processes of correlation data items are listed as follows:

[0051] (1) Volatility and correlation coefficient of two assets

[0052] Determine the rolling volatility σ1 of the first asset within a time length τ;

[0053] Determine the rolling volatility σ2 of the second asset within a time length τ;

[0054] Determine the rolling correlation coefficient ρ of the first asset and the second asset within a time length τ;

[0055] Determine the integrated volatility σ of the first asset and the second asset a The integrated volatility can be determined based on the rolling volatilities σ1, σ2 and the rolling correlation coefficient ρ. The rolling volatility can make the volatility more accurately reflect the actual situation of the market, because it can capture the changes of the market volatility, and the integrated volatility integrates the volatilities of two different assets together, so as to measure the overall volatility after their combination.

[0056] (2) determining the option price of the first asset on the second asset

[0057] The option price reflects the relationship between the price of one asset and another asset at a future time point, and the option pricing process also takes into account the volatility and correlation coefficient between the two assets. Therefore, based on the rolling volatility, correlation coefficient and integrated volatility, the option pricing formula related to the asset characteristics and the option can be used to determine the option price of the first asset on the second asset, for example, by using the two intermediate variables d1 and d2 of the normal distribution cumulative function in the option pricing formula;

[0058] Determine the normal distribution cumulative integral function N(d1), N(d2) in the option pricing formula, where the cumulative integral function values N(d1) and N(d2) are the integral values of the normal distribution density function from negative infinity to d1 and d2, and so on.

[0059] After obtaining the correlation data of the candidate assets and the target asset, the degree of correlation between the candidate assets and the target asset can be determined based on the correlation data. For example, one or more data items in the correlation data set can be analyzed to determine the criteria for distinguishing between high correlation and low correlation. For example, a data value threshold can be set for a single data item, and assets with a value higher than the threshold are considered to have high correlation. Alternatively, each single data item can be scored according to a distribution interval or a ranking interval, and the scores of multiple data items can be weighted to determine a score threshold based on the weighted combined result, and assets with a score higher than the threshold are considered to have high correlation.

[0060] In addition, according to an optional embodiment of the present disclosure, machine learning techniques can also be used in the process of determining the correlation between the candidate assets and the target asset. For example, multiple groups of assets determined to have high correlation with each other and multiple groups of assets determined to have low correlation can be used as samples, and the correlation data described above can be generated based on the asset data of each group of assets and used as training data to train a classifier model, so that the classifier model can output the correlation evaluation result of two groups of assets, i.e. high correlation or low correlation, according to the input correlation data set. It can be understood that although the above classification is binary, i.e. only two categories of high correlation and low correlation, multiple classification can also be used to train the model according to needs, so as to obtain a more fine-grained correlation evaluation result. In addition, classification also includes stratification, and accordingly, classification can include two stratification or multiple stratification.

[0061] At step 306, the sensitivity of the target asset to the price fluctuation of the candidate assets is determined for the determined candidate assets with high correlation. According to the example method described above, the assets in which the target asset is classified to present high correlation with the target asset can be determined according to the correlation data set associated with each candidate asset, and the sensitivity of the target asset to the price fluctuation of these candidate assets is further determined only in these candidate assets with high correlation. This process is equivalent to a round of screening for the candidate assets initially selected based on experience or common sense, which helps to reduce the consumption of computing resources and save the time spent in the entire process. After this round of screening, there can still be a number of candidate assets with high correlation. Therefore, the sensitivity of the price of the target asset to the price fluctuation of the remaining candidate assets, also known as the price sensitivity, can be further analyzed.

[0062] According to an embodiment of the present disclosure, the sensitivity of the target asset to the price fluctuation of the candidate assets can be determined based on a sensitivity data set. The sensitivity data set can include data reflecting the sensitivity or degree of sensitivity of the price of one asset to the price of another asset. The data in the sensitivity data set can include, but is not limited to, option sensitivity, price sensitivity, arithmetic mean of price sensitivity, geometric mean of price sensitivity, and the like. The data items in the sensitivity data set can be generated by processing the aforementioned correlation data set.

[0063] As an example, but not limitation, the following lists some example acquisition processes of sensitivity data items. The sensitivity of the option price refers to the change of the option price when a certain factor affecting the option price changes by one unit, while other factors remain unchanged, which is called the sensitivity to the factor. The sensitivity indicators of the option to different influencing factors are usually represented by five Greek letters: Delta (Δ), Gamma (Γ), Theta (Θ), Vega (ν) and Rho (ρ). The following takes the Delta (Δ) indicator as an example.

[0064] (1) Option sensitivity

[0065] According to the pricing formula of the related options, two price sensitivities of the first asset to the option price of the second asset, i.e., two deltas Δ 11 and Δ 12 , can be determined. The option sensitivity can reflect the volatility of the option price of the two assets to each other, and the higher the sensitivity, the stronger the response to the change, and the greater the volatility. Under the same hypothetical scenario, two deltas Δ and Δ of another option can also be determined.

[0066] (2) Price sensitivity

[0067] As an example, assume that two sets of option prices are equalized to obtain a relationship between the two asset prices, and thus determine the price sensitivity of the first asset to the second asset S 12 :

[0068]

[0069] (3) Arithmetic mean of price sensitivity

[0070] Determine the arithmetic mean of the price sensitivity of the first asset to the second asset for a plurality of different expiration times, for example, the sum of the price sensitivity of the first asset to the second asset for expiration times of 10, 15, 20, 25, 30, 35, 40, and 45 trading days divided by 8.

[0071] (4) Geometric mean of price sensitivity

[0072] Determine the geometric mean of the price sensitivity of the first asset to the second asset for a plurality of different expiration times, for example, the eighth root of the product of the price sensitivity of the first asset to the second asset for expiration times of 10, 15, 20, 25, 30, 35, 40, and 45 trading days.

[0073] Similarly to the aforementioned determination of correlation, the degree of sensitivity of asset price fluctuations of a target asset to a candidate asset can also be determined based on the sensitivity data. For example, a standard for distinguishing between strong sensitivity and weak sensitivity can be similarly determined based on one or more data items in the sensitivity data set. For example, by setting a threshold based on a single data item or a weighted evaluation threshold based on multiple data items, it can be determined which assets are highly sensitive to changes in their price fluctuations.

[0074] Similarly, according to an optional embodiment of the present disclosure, machine learning techniques can also be utilized to train a classification model (including binary and multi-classification models) in the process of determining the sensitivity of asset price fluctuations of a target asset to a candidate asset, so that the classifier model can output a sensitivity evaluation result of one asset to the asset price fluctuations of another asset according to the input sensitivity data set.

[0075] At step 308, the determined strong sensitivity candidate asset is selected as an anchor asset for predicting the asset price of the target asset based on the anchor asset. According to the example method described above, the candidate asset that exhibits strong sensitivity following behavior in its price fluctuation has been filtered out. Based on this, if only one strong sensitivity candidate asset is determined, the candidate asset can be selected as the anchor asset. If there are still multiple strong sensitivity candidate assets remaining, according to an embodiment of the present disclosure, these candidate assets can be further filtered based on an anchor asset dataset. The anchor asset dataset can include data reflecting the relationship between the sensitivity of an asset and the properties of the asset itself (e.g., systematic risk). The data in the anchor asset dataset can include, but is not limited to, the price sensitivity skew of one asset to another asset, the price sensitivity leverage skew of one asset to another asset, the partial derivative of price sensitivity to asset systematic risk beta, and the like. The data items in the anchor asset dataset can be generated by processing the aforementioned sensitivity dataset.

[0076] As an example, but not a limitation, some example processes for obtaining anchor asset data items are listed below:

[0077] (1) Price sensitivity skew

[0078] Cubing the sum of the corresponding price sensitivity of the first asset to the second asset at multiple different expiration times minus the corresponding arithmetic mean, and then cubing again;

[0079] (2) Price sensitivity leverage skew

[0080] Determining the product of the skew of the price of the first asset to the second asset and the corresponding geometric mean;

[0081] (3) Partial derivative of price sensitivity to asset systematic risk

[0082] Determining the partial derivative of the price sensitivity of the first asset to the second asset to asset systematic risk.

[0083] Based on the obtained anchor asset dataset, which candidate asset can be finally determined as the anchor asset of the target asset. Subsequently, any existing or other suitable asset price prediction method can be applied to predict the future price of the target asset based on the latest price data of the anchor asset. Regardless of which asset price prediction method is used, the accuracy and reliability of asset price prediction will be improved under the effect of the more accurate anchor asset selection of the present disclosure.

[0084] In yet another embodiment of the present disclosure, the anchor asset can be one asset or a group of assets, i.e., an anchor portfolio. For example, through the above process, a plurality of assets can be finally determined that have both high correlation and high sensitivity to price fluctuation of the target asset, and the target asset can be jointly influenced by these assets, and the price of the target asset can be predicted based on the combination of the price changes of these assets. For example, based on the process described above for selecting a single candidate asset as an anchor asset, the weight of each asset in the anchor portfolio in predicting the price of the target asset can be determined, and the final predicted price of the target asset can be the weighted predicted price.

[0085] Figure 4 is a flow chart illustrating an anchor-based asset price prediction method 400 according to an embodiment of the present disclosure. As shown in Figure 4 The method 400 starts with step 402, receiving historical price data of the target asset and candidate assets in a predetermined period. Then, in step 404, the correlation between the candidate assets and the target asset is determined based on the historical price data of the target asset and the candidate assets. In step 406, the sensitivity of the target asset to the asset price fluctuation of the candidate assets is determined for the candidate assets with high correlation. Finally, in step 408, the candidate assets with high sensitivity are selected as anchor assets, and the asset price of the target asset is predicted based on the anchor assets.

[0086] Figure 5 is a schematic diagram illustrating an anchor selection and anchor-based asset price prediction process according to an embodiment of the present disclosure. As shown in Figure 5 After receiving the historical price data of the target asset and the candidate assets, the asset data is first processed. Then, the correlation between the target asset and the candidate assets is determined based on the historical data, and the candidate assets with high correlation are selected in the first round of screening based on the correlation.

[0087] Then, in the candidate assets with high correlation, the sensitivity of the target asset to the asset price fluctuation of these assets is further determined, and the candidate assets with high sensitivity are selected in the second round of screening based on the sensitivity.

[0088] Finally, one or more suitable assets are selected as anchor assets as needed, and the asset price of the target asset is predicted based on the latest asset price data of the anchor assets.

[0089] Figure 6 is a data flow diagram illustrating an anchor selection and anchor-based asset price prediction process according to an embodiment of the present disclosure.

[0090] As Figure 6As shown in FIG. 1, the prediction of asset prices and the selection of anchor assets in the present disclosure both rely only on the processing of historical real asset price data. This process begins with the acquisition of historical price data of target assets and candidate assets, which can constitute an asset dataset after data preprocessing. The number of candidate assets can be determined according to the needs of the application scenario. The data items in the asset dataset can include but are not limited to: asset prices, price ratios, asset price volatility, annualized mean square deviation, normal distribution, net asset return rate variance, etc.

[0091] Subsequently, by processing the data in the asset dataset, a correlation dataset can be obtained, and a first round of screening can be performed based on the correlation. The data in the correlation dataset can include but are not limited to: the volatility and correlation coefficient of two assets, the integrated volatility of two assets, the option price of two assets, the integrated mean square deviation of two assets, the net mean square deviation of two assets with respect to the integrated mean square deviation, etc. From the correlation dataset, the number of candidate assets and the size of the associated dataset can be significantly reduced. Figure 6

[0092] Then, by processing the data in the correlation dataset, a sensitivity dataset can be further obtained, and a second round of screening can be performed based on the sensitivity. The data in the sensitivity dataset can include but are not limited to: option sensitivity, price sensitivity, arithmetic mean of price sensitivity, set mean of price sensitivity, etc. Through the sensitivity screening, the number of candidate assets and the size of the associated dataset can be further reduced.

[0093] Finally, according to the needs, the further reduced candidate asset dataset can be processed to obtain an anchor asset dataset, and one or more assets as anchor assets can be determined based on the anchor asset dataset.

[0094] In the anchor-based asset price prediction process of the present disclosure, the amount of data required for processing starts from the relatively large asset dataset of target assets and candidate assets, and gradually narrows down through correlation screening, sensitivity screening, and anchor asset data screening, finally landing on the dataset of one or more anchor assets with the strongest correlation and sensitivity, and further performing prediction, so that the resources required for prediction are relatively saved, and the efficiency is significantly improved.

[0095] ​Thus, the present disclosure solves the problem of how to determine which asset or assets have the greatest impact on the target asset, i.e., how to determine the anchor assets to be used as the basis for predicting the price changes of the target asset from among a large number of assets. At the same time, as can be seen from the above data flow diagram, the present disclosure can significantly improve the efficiency of data analysis and save the consumption of computing resources by using a plurality of data processing links and technical means such as machine learning, while ensuring the accuracy of the scheme. Thus, when applied to various asset price prediction algorithms that need to determine anchor assets, the process of determining anchor assets can be completed in a short time, thereby meeting the high requirements of the financial market for timeliness.

[0096] Figure 7 is a block diagram illustrating an anchor-based asset price prediction system 700 according to an embodiment of the present disclosure. As shown in Figure 7 The anchor-based asset price prediction system 700 can include a receiving module 702, a correlation determination module 704, a sensitivity determination module 706, and an anchor selection module 708.

[0097] The receiving module 702 can be configured to receive historical price data of a target asset and candidate assets within a predetermined period, and can preprocess the received data to form an asset data set.

[0098] The correlation determination module 704 can be configured to determine the correlation of the candidate assets with the target asset based on the historical price data of the target asset and the candidate assets. For example, the correlation determination module 704 can process the asset data set formed by the historical price data of the target asset and the candidate assets received by the receiving module 702 to obtain a correlation data set, and further determine the assets among the candidate assets that exhibit high correlation with the target asset.

[0099] The sensitivity determination module 706 can be configured to determine the sensitivity of the target asset to the asset price fluctuations of the candidate assets determined to have high correlation. For example, the sensitivity determination module 706 can process the correlation data of the candidate assets determined to have high correlation to obtain a sensitivity data set, and further determine the assets among the candidate assets that exhibit high price sensitivity.

[0100] The anchor selection module 708 can be configured to select the candidate assets determined to have high sensitivity as anchor assets for predicting the asset price of the target asset based on the anchor assets. As mentioned earlier, the anchor selection module 708 can obtain an anchor asset data set based on the processing of the sensitivity data, and select one or more assets as the final anchor assets based on these data.

[0101] Figure 8is a block diagram illustrating an anchor-based asset price prediction system 800 according to an embodiment of the present disclosure. As shown in Figure 8 The anchor-based asset price prediction system 800 can include a receiving module 802, a correlation determining module 804, a sensitivity determining module 806, an anchor selection module 808, and a prediction module 810, as shown in

[0102] The receiving module 802 can be configured to receive historical price data of target assets and candidate assets within a predetermined period, and can pre-process the received data to form an asset data set.

[0103] The correlation determining module 804 can be configured to determine the correlation of candidate assets with target assets based on the historical price data of target assets and candidate assets. For example, the correlation determining module 804 can process the asset data set formed by the historical price data of target assets and candidate assets received by the receiving module 802 to obtain a correlation data set, and further determine assets in the candidate assets that exhibit high correlation with respect to the target assets.

[0104] The sensitivity determining module 806 can be configured to determine the sensitivity of target assets to asset price fluctuations of candidate assets for the determined candidate assets with high correlation. For example, the sensitivity determining module 806 can obtain a sensitivity data set by processing the correlation data of the candidate assets determined to have high correlation, and further determine assets in the candidate assets that exhibit high price sensitivity.

[0105] The anchor selection module 808 can be configured to select the determined candidate assets with strong sensitivity as anchor assets. As mentioned earlier, the anchor selection module 808 can obtain an anchor asset data set based on processing of the sensitivity data, and select one or more assets as final anchor assets based on these data.

[0106] The prediction module 810 can be configured to predict the asset price of the target assets based on the anchor assets. The prediction module 810 can integrate one or more trained prediction models or algorithms that can be used to predict the future price of the target assets according to the input of the latest asset price or asset prices of the reference assets (i.e., the anchor assets of the present disclosure).

[0107] The various steps and modules of the anchor-based asset price prediction method and system described above can be implemented in hardware, software, or a combination thereof. If implemented in hardware, the various illustrative steps, modules, and circuits described in connection with the present disclosure can be implemented or performed with a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic component, hardware component, or any combination thereof. The general purpose processor can be a microprocessor, a controller, a microcontroller, or a state machine, etc. If implemented in software, the various illustrative steps, modules, and circuits described in connection with the present disclosure can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. The instructions or code, when executed by a processor, can cause a machine to perform the steps described in connection with the anchor-based asset price prediction method.

[0108] What has been described above includes examples of aspects of the claimed subject matter. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the claimed subject matter, but one of ordinary skill in the art will recognize that many further combinations and permutations of the claimed subject matter are possible. Accordingly, the disclosed subject matter is intended to embrace all such alterations, modifications, and variations that fall within the spirit and scope of the appended claims.

Claims

1. A method for selecting anchor points for asset price forecasting, comprising: Receive historical price data for target and candidate assets within a predetermined period; The correlation between the candidate asset and the target asset is determined based on the historical price data of the target asset and the candidate asset. For the candidate assets with high correlation, determine the sensitivity of the target asset to the price fluctuations of the candidate assets; as well as The identified highly sensitive candidate assets are used as anchor assets to predict the asset price of the target asset based on the anchor assets.

2. The method as described in claim 1, characterized in that, The historical price data of the target assets and candidate assets received within the predetermined period constitute the asset dataset.

3. The method as described in claim 2, characterized in that, The determination of the correlation between the candidate asset and the target asset is based on a correlation dataset, which is generated by processing the asset dataset.

4. The method as described in claim 3, characterized in that, The sensitivity of the target asset to the price fluctuations of the candidate assets is determined based on a sensitivity dataset, which is generated by processing the correlation dataset.

5. The method as described in claim 4, characterized in that, The process further includes using the identified highly sensitive candidate assets as anchor assets, and further filtering the identified highly sensitive candidate assets based on the anchor asset dataset, which is generated by processing the sensitivity dataset.

6. The method as described in claim 3, characterized in that, Determining the relevance between the candidate asset and the target asset includes classifying the candidate asset based on the relevance dataset, wherein the classification is binary, multi-class, two-level, or multi-level.

7. The method as described in claim 4, characterized in that, Determining the sensitivity of the target asset to the asset price fluctuations of the candidate assets includes further classifying the candidate assets based on a sensitivity dataset, wherein the classification is binary, multi-class, two-level, or multi-level.

8. The method as described in claim 1, characterized in that, The predetermined period is selected on demand.

9. The method as described in claim 1, characterized in that, The anchor asset can be a single asset or a combination of multiple assets.

10. An asset price prediction method based on anchor point selection, comprising: Anchor assets for the target asset are determined by the method described in any one of claims 1-9; as well as The asset price of the target asset is predicted based on the anchor asset.

11. An anchor point selection system for asset price forecasting, comprising: The receiving module receives historical price data of target assets and candidate assets within a predetermined period. The correlation determination module determines the correlation between the candidate asset and the target asset based on the historical price data of the target asset and the candidate asset. The sensitivity determination module determines the price fluctuation sensitivity of the target asset to the candidate assets with high correlation. as well as The anchor selection module uses the identified highly sensitive candidate assets as anchor assets to predict the asset price of the target asset based on the anchor assets.

12. An anchor-based asset price prediction system, comprising: The receiving module receives historical price data of target assets and candidate assets within a predetermined period. The correlation determination module determines the correlation between the candidate asset and the target asset based on the historical price data of the target asset and the candidate asset. The sensitivity determination module determines the price fluctuation sensitivity of the target asset to the candidate assets with high correlation. as well as The anchor point selection module selects the candidate assets with high sensitivity as anchor assets. as well as The prediction module predicts the asset price of the target asset based on the anchor asset.

13. A computer-readable storage medium having stored thereon computer instructions that, when executed, implement the method as described in any one of claims 1-10.

Images