Intelligent fund investment suggestion generation method and system

Abstract

The invention relates to the technical field of financial science and technology, and discloses an intelligent fund investment suggestion generation method and system, and the method comprises the steps: decomposing a multi-dimensional space signal into a high-frequency resonance factor and a low-frequency trend factor, and respectively constructing a dynamic resonance spectrum and a steady-state trend channel of the related emotion pulse intensity of a target investment space; calculating a risk resonance attenuation coefficient of the user investment combination; calculating a combined value drift interval of the high-frequency resonance factor, calculating a combined structure offset of the low-frequency trend factor, testing the structure offset tolerance corresponding to each strategy in the candidate balance strategy set, and determining an effective balance strategy in the candidate balance strategy set; and constructing a space risk early warning mechanism of the target investment space, performing strategy correction on the effective balance strategy to obtain a corrected balance strategy, and converting the corrected balance strategy into a fund investment suggestion of the target investment space. According to the invention, the fund investment suggestion generation accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to a method and system for generating intelligent fund investment advice, and belongs to the technical field of financial technology. Background Art

[0002] In the field of financial investment, funds, as a form of collective investment, have been favored by the majority of investors due to their professional management and risk diversification. With the development of the economy and the continuous innovation of the financial market, the types of fund products have become increasingly rich, from traditional stock and bond funds to diversified categories such as index funds, quantitative funds, and cross-border funds. Faced with such a huge fund market, investors face huge challenges in choosing suitable funds for investment. At the same time, different investors have significant differences in investment goals, risk tolerance, financial status, and investment experience, which makes the demand for personalized and precise fund investment advice more urgent.

[0003] At present, there are many problems with the way fund investment advice is generated. Traditional manual investment advice mainly relies on the personal experience and professional knowledge of investment advisors. Investment advisors provide investment advice to investors based on their own judgment of the market, understanding of fund products and past investment cases. However, this method has obvious limitations. On the one hand, investment advisors have limited time and energy, and it is difficult to provide timely and comprehensive services to a large number of investors at the same time. On the other hand, the subjective judgment of investment advisors is easily affected by personal experience and emotions, resulting in difficulty in ensuring the accuracy and stability of investment advice. In addition, the market environment changes rapidly, and it is difficult for investment advisors to track the dynamic information of all fund products in real time, which makes it possible for investment advice to fail to adapt to market changes in a timely manner, thereby resulting in low accuracy in the generation of fund investment advice. Summary of the Invention

[0004] The present invention provides an intelligent fund investment advice generation method and system, the main purpose of which is to improve the accuracy of generating fund investment advice.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for generating intelligent fund investment advice, comprising:

[0006] Extracting the associated fund clusters of the target investment space from a pre-built cross-space investment database, analyzing the risk-return constraints corresponding to the target investment space, capturing the multidimensional spatial signals of the associated fund clusters in real time, decomposing the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and constructing a dynamic resonance spectrum and a steady-state trend channel related to the emotional pulse intensity of the target investment space;

[0007] Collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient of the user investment combination;

[0008] Based on the dynamic resonance graph and the resource fluctuation transmission network, a two-way impact model of the target input space is constructed. Based on the two-way impact model, the combined value drift interval of the high-frequency resonance factor is calculated, and the combined structural offset of the low-frequency trend factor is calculated in combination with the steady-state trend channel;

[0009] generating a candidate balancing strategy set corresponding to the portfolio value drift interval according to the risk-return constraint condition, testing the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural deviation, and determining an effective balancing strategy in the candidate balancing strategy set based on the structural deviation tolerance;

[0010] Combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, a spatial risk warning mechanism for the target investment space is constructed. Based on the spatial risk warning mechanism, the effective balance strategy is revised to obtain a revised balance strategy, and the revised balance strategy is converted into a fund investment recommendation for the target investment space.

[0011] Optionally, analyzing the risk-return constraint conditions corresponding to the target investment space includes:

[0012] Collecting spatial ecological panoramic data corresponding to the target input space, and performing data cleaning on the spatial ecological panoramic data to obtain target ecological panoramic data;

[0013] Performing structured processing on the target ecological panoramic data to obtain structured ecological panoramic data;

[0014] Extracting eco-labels corresponding to the structural ecological panorama data, and analyzing risk-benefit labels from the eco-labels;

[0015] Based on the risk-benefit label, extracting risk-benefit characteristics from the structural ecological panorama data;

[0016] Based on the risk-return characteristics, the risk-return constraints corresponding to the target investment space are analyzed.

[0017] Optionally, decomposing the multidimensional spatial signal into a high-frequency resonance factor and a low-frequency trend factor includes:

[0018] performing signal noise reduction processing on the multidimensional spatial signal to obtain a noise-reduced multidimensional spatial signal;

[0019] Analyzing the spatial signal dimension corresponding to the denoised multidimensional spatial signal, and constructing a three-dimensional signal tensor of the denoised multidimensional spatial signal based on the spatial signal dimension;

[0020] Performing missing value completion processing on the signal three-dimensional tensor to obtain a target three-dimensional tensor;

[0021] Performing matrix decomposition on the target three-dimensional tensor to obtain a basis matrix and a coefficient matrix;

[0022] A high-frequency resonance factor and a low-frequency trend factor of the multi-dimensional spatial signal are extracted from the base matrix and the coefficient matrix.

[0023] Optionally, combining the resource distribution matrix and the flow characteristic parameters and using a trained deep learning time series prediction model to generate a resource fluctuation transmission network under different spatial scenarios in the target input space includes:

[0024] Using the input layer in the deep learning time series prediction model to adjust the resource distribution matrix and the flow characteristic parameters respectively, to obtain an adjusted resource distribution matrix and an adjusted flow characteristic parameter;

[0025] Calculating the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters using the self-attention function in the deep learning time series prediction model;

[0026] Utilizing the time series convolution layer in the deep learning time series prediction model, respectively extracting the time series features corresponding to the adjusted resource distribution matrix and the adjusted flow characteristic parameters to obtain matrix time series features and parameter time series features;

[0027] Analyzing the timing dependency between the matrix timing features and the parameter timing features using the timing cycle layer in the deep learning timing prediction model;

[0028] Combining the spatiotemporal coupling coefficient and the temporal dependency, and utilizing a fully connected layer in the deep learning temporal prediction model to perform feature fusion on the matrix temporal features and the parameter temporal features to obtain a comprehensive temporal feature;

[0029] Based on the comprehensive time series features, the network topology optimization unit in the deep learning time series prediction model is used to generate a resource fluctuation transmission network under different spatial scenarios in the target input space.

[0030] Optionally, the calculating the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters by using the self-attention function in the deep learning time series prediction model includes:

[0031] Projecting the adjusted resource distribution matrix onto the corresponding query space to obtain a resource query vector;

[0032] Projecting the adjusted flow characteristic parameters onto a corresponding key space to obtain a parameter key vector;

[0033] Combining the resource query vector and the parameter key vector, the self-attention function is used to calculate the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters:

[0034] The self-attention function calculation formula is as follows:

[0035]

[0036] Where A represents the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters, B(a) represents the resource query vector corresponding to the a-th matrix in the adjustment resource distribution matrix, D(b) represents the parameter key vector corresponding to the b-th parameter in the adjustment flow characteristic parameters, a represents the sequence number of the adjustment resource distribution matrix, b represents the sequence number of the adjustment flow characteristic parameters, and β represents the scaling factor.

[0037] Optionally, the calculating the risk resonance attenuation coefficient of the user investment combination includes:

[0038] Collecting historical rate of return data of the user's investment portfolio, and constructing a rate of return matrix of the user's investment portfolio based on the historical rate of return data;

[0039] Calculating the covariance matrix corresponding to the yield matrix, performing dynamic conditional correlation coefficient decomposition on the covariance matrix to obtain a time-varying correlation coefficient matrix;

[0040] Determining a risk resonance event of the user investment combination based on the time-varying correlation coefficient matrix;

[0041] Calculating the risk resonance intensity corresponding to the risk resonance event, fitting the attenuation factor corresponding to the risk resonance intensity, and defining the half-life corresponding to the risk resonance intensity;

[0042] Combining the attenuation factor and the half-life, the risk resonance attenuation coefficient of the user investment portfolio is calculated using the following formula:

[0043]

[0044] Among them, F represents the risk resonance attenuation coefficient of the user's investment portfolio, λ represents the attenuation factor, and T half Indicates half-life.

[0045] Optionally, constructing a bidirectional impact model of the target input space based on the dynamic resonance spectrum and the resource fluctuation transmission network includes:

[0046] extracting a dynamic resonance feature corresponding to the dynamic resonance spectrum, and performing quantization processing on the dynamic resonance feature to obtain a dynamic resonance feature vector;

[0047] Performing topological analysis on the resource fluctuation transmission network to obtain network structure characteristics;

[0048] Performing parameter conversion on the network structure characteristics to obtain a resource fluctuation network parameter set;

[0049] Combining the dynamic resonance eigenvector and the resource fluctuation network parameter set, a bidirectional impact model of the target input space is constructed.

[0050] Optionally, calculating the combined structural offset of the low-frequency trend factor in combination with the steady-state trend channel includes:

[0051] Calculating a steady-state trend constraint parameter in the steady-state trend channel, and extracting a net value deviation signal corresponding to the steady-state trend constraint parameter;

[0052] Identifying a signal sequence corresponding to the net value deviation signal, and calculating a structural deviation intensity corresponding to the net value deviation signal based on the signal sequence;

[0053] The market benchmark offset strength corresponding to the steady-state trend channel is queried, and the combined structural offset of the low-frequency trend factor is calculated by combining the market benchmark strength and the structural deviation strength using the following formula:

[0054]

[0055] Among them, δ represents the combined structural offset of the low-frequency trend factor, G adj Indicates the structural deviation strength, G benchmark Indicates the market benchmark deviation strength, max(G adj ,G benchmark ) means selecting the maximum value between the market benchmark strength and the structural deviation strength.

[0056] Optionally, the testing of the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the combined structural deviation includes:

[0057] Performing scene construction processing on the combined structure offset to obtain a structure offset scene;

[0058] Performing a comprehensive simulation process on each strategy in the candidate balancing strategy set and the structural shift scenario, and recording dynamic data of investment indicators during the comprehensive simulation process;

[0059] Identifying a risk-return indicator in the investment indicator dynamic data, and calculating a tolerance score corresponding to the risk-return indicator based on the investment indicator dynamic data;

[0060] Allocating a return indicator weight corresponding to the risk-return indicator, and calculating a comprehensive tolerance score corresponding to each strategy in the candidate balancing strategy set by combining the return indicator weight and the tolerance score;

[0061] Based on the comprehensive tolerance score, the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set is evaluated.

[0062] In order to solve the above problems, the present invention also provides an intelligent fund investment advice generation system, which includes:

[0063] A trend channel construction module is used to extract the associated fund clusters of the target investment space from a pre-built cross-space investment database, analyze the risk-return constraints corresponding to the target investment space, capture the multidimensional spatial signals of the associated fund clusters in real time, decompose the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and respectively construct a dynamic resonance spectrum and a steady-state trend channel related to the emotional pulse intensity of the target investment space;

[0064] A risk resonance attenuation coefficient calculation module is used to collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate a resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient of the user investment combination;

[0065] a combined structure offset calculation module, configured to construct a two-way impact model of the target input space based on the dynamic resonance map and the resource fluctuation transmission network, calculate the combined value drift interval of the high-frequency resonance factor based on the two-way impact model, and calculate the combined structure offset of the low-frequency trend factor in combination with the steady-state trend channel;

[0066] an effective balancing strategy determination module, configured to generate a set of candidate balancing strategies corresponding to the portfolio value drift interval based on the risk-return constraint condition, test the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural deviation, and determine an effective balancing strategy in the candidate balancing strategy set based on the structural deviation tolerance;

[0067] A recommendation generation module is used to combine the liquidity characteristic parameters with the risk resonance attenuation coefficient to construct a spatial risk warning mechanism for the target investment space, and based on the spatial risk warning mechanism, to modify the effective balance strategy to obtain a modified balance strategy, and convert the modified balance strategy into a fund investment recommendation for the target investment space.

[0068] Compared with the problems described in the background technology, the present invention extracts the associated fund clusters of the target investment space from the pre-constructed cross-space investment database, and analyzes the risk-return constraints corresponding to the target investment space, which can provide insights into the capital distribution and potential risk-return status of the target investment space, thereby laying an important foundation for the subsequent generation of the candidate balance strategy set corresponding to the portfolio value drift interval. Furthermore, the present invention combines the resource distribution matrix and the flow characteristic parameters, and uses the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, which can clearly present the propagation path and mutual influence relationship of resource fluctuations in different spatial scenarios, so as to improve the accuracy of the subsequent construction of the two-way impact model of the target investment space. The present invention constructs a two-way impact model based on the dynamic resonance spectrum and the resource fluctuation transmission network, which can comprehensively consider the interaction of multiple complex factors in the market, and improve the subsequent generation of candidate balance strategies corresponding to the portfolio value drift interval. The accuracy of the balance strategy set is improved. Furthermore, the present invention generates a candidate balance strategy set corresponding to the portfolio value drift interval based on the risk-return constraint conditions, which can provide investors with diversified potential investment strategy options that are in line with their risk-return demands, helping investors to flexibly layout in a complex and changing market environment. Based on the portfolio structure offset, the structural deviation tolerance corresponding to each strategy in the candidate balance strategy set is tested, which can accurately evaluate the ability of each strategy to maintain stable returns when responding to changes in the portfolio structure, and provide a strong basis for investors to screen out more adaptable and reliable investment strategies. Furthermore, the present invention constructs a spatial risk warning mechanism for the target investment space by combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, comprehensively considering the factors related to capital flow and risk, which can significantly improve the accuracy of risk warning, and based on the spatial risk warning mechanism, the effective balance strategy is revised, thereby enhancing the adaptability and robustness of the fund investment strategy. Therefore, the intelligent fund investment advice generation method and system provided by the embodiment of the present invention can improve the accuracy of fund investment advice generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A flowchart of a method for generating intelligent fund investment advice according to an embodiment of the present invention;

[0070] Figure 2A schematic diagram of modules for implementing the intelligent fund investment advice generation method provided in one embodiment of the present invention.

[0071] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0073] Embodiments of the present application provide a method for generating intelligent fund investment advice. The method may be executed by at least one of a server, a terminal, or other electronic device capable of executing the method provided by the embodiments of the present application. In other words, the method may be executed by software or hardware installed on a terminal device or a server device. The server may include, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0074] Example 1:

[0075] Reference Figure 1 FIG. 1 is a flow chart of a method for generating smart fund investment advice according to an embodiment of the present invention. In this embodiment, the method for generating smart fund investment advice includes:

[0076] S1. Extract the associated fund clusters of the target investment space from the pre-built cross-space investment database, analyze the risk-return constraints corresponding to the target investment space, capture the multidimensional spatial signals of the associated fund clusters in real time, decompose the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and respectively construct the dynamic resonance spectrum and steady-state trend channel of the target investment space related to the emotional pulse intensity.

[0077] The present invention extracts the associated fund clusters of the target investment space from the pre-constructed cross-space investment database, and analyzes the risk-return constraints corresponding to the target investment space. This can provide insights into the capital distribution and potential risk-return status of the target investment space, thereby laying an important foundation for the subsequent generation of the candidate equilibrium strategy set corresponding to the portfolio value drift interval. It should be explained that the pre-constructed cross-space investment database is a huge investment information treasure house, which brings together fund data from different industries, regions and multiple investment strategies. The target investment space is a specific investment category focused on by investors based on specific investment concepts, market trend judgments or industry research interests, such as the popular artificial intelligence industry investment space, or the emerging energy regional investment space that a country focuses on supporting. The risk-return constraints are a series of restrictions and guiding factors corresponding to the target investment space, covering the market volatility tolerance range, capital liquidity requirements, expected investment return requirements, etc., which are used to define the trade-off between risk and return in this investment space.

[0078] Specifically, the analysis of the risk-return constraints corresponding to the target investment space includes:

[0079] Collecting spatial ecological panoramic data corresponding to the target input space, and performing data cleaning on the spatial ecological panoramic data to obtain target ecological panoramic data;

[0080] Performing structured processing on the target ecological panoramic data to obtain structured ecological panoramic data;

[0081] Extracting eco-labels corresponding to the structural ecological panorama data, and analyzing risk-benefit labels from the eco-labels;

[0082] Based on the risk-benefit label, extracting risk-benefit characteristics from the structural ecological panorama data;

[0083] Based on the risk-return characteristics, the risk-return constraints corresponding to the target investment space are analyzed.

[0084] It should be explained that the spatial ecological panoramic data is a comprehensive data set corresponding to the target investment space, covering multi-dimensional investment environment and trends such as macroeconomics, industry dynamics, and market history; the target ecological panoramic data is the data of the spatial ecological panoramic data after data cleaning to remove interference information such as noise and outliers; the structural ecological panoramic data is the data of the target ecological panoramic data after extracting key information and converting it into a structured format; the ecological label is the identifier corresponding to the structural ecological panoramic data that classifies and labels the data based on the key dimensions of investment analysis; the risk-return label is a specific classification identifier in the ecological label that is directly related to the investment risk and return assessment; the risk-return characteristics are the key information in the data subset corresponding to the risk-return label in the structural ecological panoramic data that can reflect the investment risk and return characteristics.

[0085] Furthermore, the spatial ecological panoramic data corresponding to the target investment space can be collected through a professional financial data platform, and the spatial ecological panoramic data can be cleaned by a box plot method to obtain target ecological panoramic data; the target ecological panoramic data can be structured by a natural language processing tool to obtain structural ecological panoramic data; the ecological labels corresponding to the structural ecological panoramic data can be extracted by a data annotation tool, and the data annotation tool is compiled by a scripting language; the risk-return label can be analyzed from the ecological label by a keyword matching algorithm; based on the risk-return label, the risk-return characteristics can be extracted from the structural ecological panoramic data, and the return characteristics can be extracted by calculating indicators such as historical yield, expected yield, and return fluctuation range, and the risk characteristics can be extracted by calculating indicators such as volatility, maximum drawdown, and downside risk; based on the risk-return characteristics, by comparing similar situations in historical data, referring to the industry's authoritative risk-return assessment standards, combining the market status and the investor's own risk tolerance, a comprehensive analysis of the risk-return constraints corresponding to the target investment space is conducted, such as evaluating the reasonable range of expected returns under different risk levels and the maximum risk level that can be tolerated under specific return expectations.

[0086] By decomposing the multidimensional spatial signal into a high-frequency resonance factor and a low-frequency trend factor, the present invention can analyze the short-term volatility characteristics and long-term development trends of the associated fund cluster, helping investors flexibly adjust their investment strategies based on market dynamics of varying frequencies in a complex and volatile market environment. It should be explained that the multidimensional spatial signal is generated during the market operation of the associated fund cluster by the interweaving and combined effects of multiple factors, including fund net value fluctuations, trading activity, capital flows, changes in investor sentiment, and macroeconomic and industry dynamics. It is a complex data set that can comprehensively reflect the operating status and market characteristics of the associated fund cluster from multiple dimensions. The high-frequency resonance factor and the low-frequency trend factor, after decomposition of the multidimensional spatial signal in the time and frequency dimensions, respectively, represent the dramatic signal fluctuation components caused by factors such as rapid changes in market sentiment and sudden fluctuations in hot spots in the short term, and the slow and continuous signal change trend components influenced by stable factors such as the macroeconomic and industry development cycles in the long term. Furthermore, the multidimensional spatial signal of the associated fund cluster can be captured in real time using an Octopus collector.

[0087] In detail, decomposing the multidimensional spatial signal into high-frequency resonance factors and low-frequency trend factors includes:

[0088] performing signal noise reduction processing on the multidimensional spatial signal to obtain a noise-reduced multidimensional spatial signal;

[0089] Analyzing the spatial signal dimension corresponding to the denoised multidimensional spatial signal, and constructing a three-dimensional signal tensor of the denoised multidimensional spatial signal based on the spatial signal dimension;

[0090] Performing missing value completion processing on the signal three-dimensional tensor to obtain a target three-dimensional tensor;

[0091] Performing matrix decomposition on the target three-dimensional tensor to obtain a basis matrix and a coefficient matrix;

[0092] A high-frequency resonance factor and a low-frequency trend factor of the multi-dimensional spatial signal are extracted from the base matrix and the coefficient matrix.

[0093] It should be explained that the denoised multidimensional spatial signal is a pure version of the multidimensional spatial signal after noise interference is removed; the spatial signal dimension is a quantitative indicator corresponding to the denoised multidimensional spatial signal that describes the changes in the signal in different directions or characteristics; the signal three-dimensional tensor is a data structure in which the denoised multidimensional spatial signal is presented in the form of a three-dimensional array, which is used to more comprehensively and three-dimensionally represent the multi-faceted information of the signal; the basis matrix and the coefficient matrix are obtained after the target three-dimensional tensor is decomposed by a specific matrix decomposition algorithm, and are used to interpret the target three-dimensional tensor data characteristics from different angles, so as to assist in the subsequent analysis of the data matrix of high-frequency resonance factors and low-frequency trend factors.

[0094] Furthermore, the multidimensional spatial signal can be subjected to signal noise reduction processing by the wavelet transform method to obtain a noise-reduced multidimensional spatial signal; the spatial signal dimension corresponding to the noise-reduced multidimensional spatial signal can be analyzed by the principal component analysis method; the signal three-dimensional tensor of the noise-reduced multidimensional spatial signal can be constructed according to the time series, different characteristic attributes and spatial distribution characteristics of the signal; the missing value completion processing of the signal three-dimensional tensor can be performed by the high-order singular value decomposition method to obtain a target three-dimensional tensor; the target three-dimensional tensor can be subjected to matrix decomposition processing by the non-negative matrix decomposition algorithm to obtain a basis matrix and a coefficient matrix; the high-frequency resonance factor and the low-frequency trend factor of the multidimensional spatial signal can be extracted from the basis matrix and the coefficient matrix by analyzing the change rules of the elements in the basis matrix and the coefficient matrix, combined with the frequency analysis method, according to the principle that high-frequency components correspond to rapid changes and low-frequency components correspond to slow trends.

[0095] The present invention can intuitively present the sharp fluctuations in market sentiment in the short term and the stable trend in the long term by respectively constructing a dynamic resonance spectrum and a steady-state trend channel of the target investment space related to the intensity of the emotion pulse, thereby providing convenience for subsequent related processing. It should be explained that the dynamic resonance spectrum and the steady-state trend channel are respectively intuitive visual expressions of the intensity of the emotion pulse in the target investment space. The former graphically displays the rapid fluctuations, mutual influence and resonance changes of the emotion pulse in the short term in real time, while the latter presents the relatively stable development situation and trend of the emotion pulse intensity in the long term. The combination of the two provides investors with a comprehensive and clear insight into market sentiment. Furthermore, by collecting investor behavior data, social media public opinion tendencies and market trading activity data in the target investment space, and using data mining algorithms and visualization technologies, the dynamic resonance spectrum and steady-state trend channel of the emotion pulse intensity in the target investment space can be respectively constructed.

[0096] S2. Collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient within the user investment combination.

[0097] The present invention combines the resource distribution matrix and the flow characteristic parameters, and uses the trained deep learning time series prediction model to generate a resource fluctuation transmission network under different spatial scenarios in the target input space. It can clearly present the propagation path and mutual influence relationship of resource fluctuations in different spatial scenarios, so as to improve the accuracy of constructing the subsequent two-way impact model of the target input space.

[0098] It should be explained that the resource distribution matrix is ​​a data set that presents the distribution status of the quantity, proportion, position and other aspects of the various resources of the user investment combination in the target investment space in the space in the form of a matrix. The flow characteristic parameters are related parameters of the speed, direction, frequency and other characteristics of the resources of the user investment combination in the target investment space when flowing in the space. The deep learning time series prediction model is a model based on deep learning architecture that has been trained with a large amount of historical data and can learn the inherent laws and characteristics of time series data, so as to predict future resource fluctuations. The resource fluctuation transmission network is an intuitive representation of how resource fluctuations under different spatial scenarios in the target investment space propagate, diffuse and influence each other among various components in the form of a network structure, reflecting the path and correlation degree of resource fluctuations. Furthermore, the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space can be collected through sensors, data acquisition systems and intelligent monitoring equipment installed at key nodes.

[0099] In detail, the resource distribution matrix and the flow characteristic parameters are combined to generate a resource fluctuation transmission network under different spatial scenarios in the target input space using a trained deep learning time series prediction model, including:

[0100] Using the input layer in the deep learning time series prediction model to adjust the resource distribution matrix and the flow characteristic parameters respectively, to obtain an adjusted resource distribution matrix and an adjusted flow characteristic parameter;

[0101] Calculating the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters using the self-attention function in the deep learning time series prediction model;

[0102] Utilizing the time series convolution layer in the deep learning time series prediction model, respectively extracting the time series features corresponding to the adjusted resource distribution matrix and the adjusted flow characteristic parameters to obtain matrix time series features and parameter time series features;

[0103] Analyzing the timing dependency between the matrix timing features and the parameter timing features using the timing cycle layer in the deep learning timing prediction model;

[0104] Combining the spatiotemporal coupling coefficient and the temporal dependency, and utilizing a fully connected layer in the deep learning temporal prediction model to perform feature fusion on the matrix temporal features and the parameter temporal features to obtain a comprehensive temporal feature;

[0105] Based on the comprehensive time series features, the network topology optimization unit in the deep learning time series prediction model is used to generate a resource fluctuation transmission network under different spatial scenarios in the target input space.

[0106] It should be explained that the input layer is the starting part of the deep learning time series prediction model that is responsible for receiving the original data and performing preliminary format adjustment and adaptation. It is composed of input neurons and data preprocessing modules, and is used to perform operations such as normalization on the input data; the adjusted resource distribution matrix and the adjusted flow characteristic parameters are respectively the resource distribution matrix and the flow characteristic parameters that are adjusted by the input layer, such as data cleaning, standardization, dimensionality transformation, etc., to obtain a data form that is more suitable for subsequent model processing; the self-attention function is a mechanism in the deep learning time series prediction model for calculating the degree of correlation between different data elements, by assigning weights to each element. Highlight the relationship between important elements; the spatiotemporal coupling coefficient is a quantitative indicator of the degree of interaction and mutual influence between the adjustment resource distribution matrix and the adjustment flow characteristic parameters in the spatial and temporal dimensions; the time series convolution layer is a part that uses the convolution operation in the deep learning time series prediction model to extract the time series data features, and is composed of a convolution kernel, an activation function, and a feature mapping calculation module, etc., and captures local features by sliding the convolution kernel on the time series; the matrix time series features and the parameter time series features are respectively extracted from the adjustment resource distribution matrix and the adjustment flow characteristic parameters after being processed by the time series convolution layer, reflecting their respective change laws and characteristics in the time series. The representation of the feature; the time series recurrent layer is the part of the deep learning time series prediction model used to process the long-term dependency of time series data, which is composed of memory units (such as cell states in LSTM) and gating units (input gate, forget gate, output gate, etc.), which can remember past information and update the state according to the current input; the time series dependency is the mutual dependence and mutual influence relationship between the matrix time series features and the parameter time series features in the time sequence, that is, the change of one feature may be affected by the past state of another feature; the fully connected layer is the deep learning time series prediction model used to integrate and transform the features extracted by the previous layers and map them to the output space The part is composed of multiple neurons, each neuron is connected to all neurons in the previous layer and performs weighted summation and activation operations; the comprehensive time series feature is the information of the matrix time series feature and the parameter time series feature obtained after a fusion operation (such as element-by-element addition, weighted merging, etc.), which can more comprehensively reflect the representation of the comprehensive characteristics of the data in the time series; the network topology optimization unit is the part of the deep learning time series prediction model used to optimize and adjust the generated network structure, which is composed of a topology analysis module, an edge and node adjustment module, and an optimization strategy formulation module, etc., and improves the network topology structure according to certain rules and goals to make it more in line with actual needs.

[0107] Furthermore, as an optional embodiment of the present invention, the calculating of the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters using the self-attention function in the deep learning time series prediction model includes:

[0108] Projecting the adjusted resource distribution matrix onto the corresponding query space to obtain a resource query vector;

[0109] Projecting the adjusted flow characteristic parameters onto a corresponding key space to obtain a parameter key vector;

[0110] Combining the resource query vector and the parameter key vector, the self-attention function is used to calculate the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters:

[0111] The self-attention function calculation formula is as follows:

[0112]

[0113] Where A represents the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters, B(a) represents the resource query vector corresponding to the a-th matrix in the adjustment resource distribution matrix, D(b) represents the parameter key vector corresponding to the b-th parameter in the adjustment flow characteristic parameters, a represents the sequence number of the adjustment resource distribution matrix, b represents the sequence number of the adjustment flow characteristic parameters, and β represents the scaling factor.

[0114] It should be explained that the resource query vector is a vector representation obtained by projecting the adjusted resource distribution matrix into the query space after a specific linear transformation, and is used to find correlations with other data elements in the self-attention mechanism; the parameter key vector is a vector form obtained by linearly mapping the adjusted flow feature parameters in the key space, which can be matched with the query vector to measure the degree of association, and the scaling factor is the square root of the dimensions of the resource query vector and the parameter key vector.

[0115] Furthermore, the adjusted resource distribution matrix can be projected onto the corresponding query space by performing matrix multiplication operations on a learnable weight matrix adapted to the query space dimension to obtain a resource query vector; the adjusted flow characteristic parameters can be projected onto the corresponding key space by performing matrix multiplication operations on a learnable weight matrix adapted to the key space dimension to obtain a parameter key vector.

[0116] By calculating the risk resonance attenuation coefficient of the user investment combination, the present invention can measure the mutual influence and dissipation degree of risks among the various resources in the user investment combination, and provide an important basis for the subsequent construction of the spatial risk early warning mechanism of the target investment space. It should be explained that the risk resonance attenuation coefficient is a quantitative indicator of the degree of risk intensity attenuation after the various resources in the user investment combination generate risk resonance due to interaction in the target investment space, as time goes by or during the risk transmission process. It reflects the closeness of the risk association between the resources in the combination and the speed and ability of the risk dissipation in the combination.

[0117] Specifically, the calculation of the risk resonance attenuation coefficient of the user's investment portfolio includes:

[0118] Collecting historical rate of return data of the user's investment portfolio, and constructing a rate of return matrix of the user's investment portfolio based on the historical rate of return data;

[0119] Calculating the covariance matrix corresponding to the yield matrix, performing dynamic conditional correlation coefficient decomposition on the covariance matrix to obtain a time-varying correlation coefficient matrix;

[0120] Determining a risk resonance event of the user investment combination based on the time-varying correlation coefficient matrix;

[0121] Calculating the risk resonance intensity corresponding to the risk resonance event, fitting the attenuation factor corresponding to the risk resonance intensity, and defining the half-life corresponding to the risk resonance intensity;

[0122] Combining the attenuation factor and the half-life, the risk resonance attenuation coefficient of the user investment portfolio is calculated using the following formula:

[0123]

[0124] Among them, F represents the risk resonance attenuation coefficient of the user's investment portfolio, λ represents the attenuation factor, and T half Indicates half-life.

[0125] It should be explained that the historical yield data is the historical yield sequence of the assets in the user's investment portfolio; the yield matrix is ​​a two-dimensional arrangement of the historical yields of multiple assets in the user's investment portfolio; the covariance matrix is ​​a quantitative matrix of the correlation between the return fluctuations of each asset corresponding to the yield matrix; the time-varying correlation coefficient matrix is ​​the time-varying correlation relationship obtained by decomposing the covariance matrix through the dynamic conditional correlation model; the risk resonance event is an extreme market event that simultaneously amplifies the risks of multiple assets in the user's investment portfolio; the risk resonance intensity is the degree of risk impact corresponding to the risk resonance event, the attenuation factor is the exponential attenuation parameter corresponding to the risk resonance intensity, and the half-life is the time required for the impact corresponding to the risk resonance intensity to decay by half.

[0126] Furthermore, the historical yield data of the user's investment portfolio can be collected through a financial data platform. Based on the historical yield data, the yield matrix of the user's investment portfolio can be constructed by arranging the yields of each asset in chronological order; the covariance matrix corresponding to the yield matrix can be calculated through the covariance formula, and the covariance matrix can be subjected to dynamic conditional correlation coefficient decomposition through the DCC-GARCH model to obtain a time-varying correlation coefficient matrix; based on the time-varying correlation coefficient matrix, the risk resonance event of the user's investment portfolio can be determined by setting a correlation coefficient threshold and traversing the matrix; the risk resonance intensity can be obtained by calculating the average correlation coefficient corresponding to the risk resonance event, the attenuation factor corresponding to the risk resonance intensity can be fitted by the least squares method combined with the data of the risk resonance intensity changing with time after the occurrence of historical risk resonance events, and the half-life corresponding to the risk resonance intensity can be defined by solving the attenuation function so that the risk resonance intensity decays to half of the time.

[0127] S3. Based on the dynamic resonance spectrum and the resource fluctuation transmission network, a two-way impact model of the target input space is constructed. Based on the two-way impact model, the combined value drift interval of the high-frequency resonance factor is calculated, and the combined structural offset of the low-frequency trend factor is calculated in combination with the steady-state trend channel.

[0128] The present invention constructs a two-way impact model based on the dynamic resonance spectrum and the resource fluctuation transmission network, which can comprehensively consider the interaction of various complex factors in the market and improve the accuracy of the subsequent generation of the candidate equilibrium strategy set corresponding to the portfolio value drift interval. It should be explained that the two-way impact model is a comprehensive quantitative analysis tool for the target investment space, integrating the key information in the dynamic resonance spectrum and the resource fluctuation transmission network, and constructing equations including asset price changes, resource flows and the interaction between the two to simulate the two-way impact of high-frequency resonance factors and low-frequency trend factors on the asset value and investment portfolio structure in the target investment space under different market conditions, providing a basis for evaluating the risk and return status of this space.

[0129] In detail, the bidirectional impact model of the target input space is constructed based on the dynamic resonance spectrum and the resource fluctuation transmission network, including:

[0130] extracting a dynamic resonance feature corresponding to the dynamic resonance spectrum, and performing quantization processing on the dynamic resonance feature to obtain a dynamic resonance feature vector;

[0131] Performing topological analysis on the resource fluctuation transmission network to obtain network structure characteristics;

[0132] Performing parameter conversion on the network structure characteristics to obtain a resource fluctuation network parameter set;

[0133] Combining the dynamic resonance eigenvector and the resource fluctuation network parameter set, a bidirectional impact model of the target input space is constructed.

[0134] It should be explained that the dynamic resonance feature is the key information embodiment corresponding to the dynamic resonance map, reflecting the specific characteristics of the resonance generated by the interaction of assets in the target investment space in the short term; the dynamic resonance feature vector is the quantitative numerical representation of the dynamic resonance feature, which facilitates subsequent mathematical calculations and analysis by converting the feature into a vector form; the network structure feature is the core component of the resource fluctuation transmission network, which describes the basic architecture and key attributes of the flow of resources in the network, such as node and path structure; the resource fluctuation network parameter set is the mathematical expression of the network structure feature, which converts the network structure feature into a computable parameter set, providing necessary data support for the construction of a two-way impact model.

[0135] Furthermore, the dynamic resonance features corresponding to the dynamic resonance spectrum can be extracted through image recognition algorithms and data analysis tools, such as identifying the price fluctuation curve trend in the spectrum, capturing the start and end time of the resonance signal and other key features; the dynamic resonance features can be quantified through mathematical quantification methods to obtain dynamic resonance feature vectors, such as converting the price fluctuation amplitude into a standardized value, counting and normalizing the resonance frequency, etc.; the resource fluctuation transmission network can be topologically analyzed through graph theory analysis software to obtain network structure features, such as identifying hub nodes in the network, determining the connection density of edges, etc.; the network structure features can be parameterized through a specific parameterization algorithm to obtain a resource fluctuation network parameter set, such as converting the node importance into a weight parameter, quantifying the connection density of edges into parameters such as the flow conduction coefficient; combining the dynamic resonance feature vector and the resource fluctuation network parameter set, constructing a two-way impact model of the target input space, combining the dynamic resonance feature vector with the resource fluctuation network parameter set, applying the system dynamics principle for modeling, and constructing a two-way impact model that includes the asset price change equation, the resource flow equation, and the interaction equation between the two.

[0136] The present invention calculates the portfolio value drift interval of the high-frequency resonance factor based on the two-way impact model, which can accurately grasp the fluctuation range of the investment portfolio value when the market fluctuates violently in the short term, laying a foundation for the generation of the subsequent candidate balance strategy set. In combination with the steady-state trend channel, the portfolio structure offset of the low-frequency trend factor is calculated, which can clearly understand the degree to which the investment portfolio structure deviates from the ideal state in long-term investment, and help investors optimize asset allocation in a timely manner according to the long-term market trend. It should be explained that the portfolio value drift interval is the high-frequency resonance factor under a certain confidence level, based on the two-way impact model to simulate different market scenarios, reflecting the portfolio value. The quantitative reflection of the fluctuation range can occur; the portfolio structure offset is the quantitative result reflecting the degree of deviation of the portfolio structure from the steady-state trend channel by defining specific indicators and calculating the deviation between the proportion of various assets in the portfolio and the ideal proportion under the steady-state trend channel when the steady-state trend channel evaluates the low-frequency trend factor. Furthermore, based on the two-way impact model, the portfolio value drift range of the high-frequency resonance factor is calculated. For example, by simulating multiple market scenarios, including extreme fluctuations and stable states, the impact of the high-frequency resonance factor on the portfolio value in different scenarios is analyzed, and statistical methods are used to calculate the fluctuation range of the portfolio value under a certain confidence level.

[0137] In detail, the calculation of the combined structural offset of the low-frequency trend factor in combination with the steady-state trend channel includes:

[0138] Calculating a steady-state trend constraint parameter in the steady-state trend channel, and extracting a net value deviation signal corresponding to the steady-state trend constraint parameter;

[0139] Identifying a signal sequence corresponding to the net value deviation signal, and calculating a structural deviation intensity corresponding to the net value deviation signal based on the signal sequence;

[0140] The market benchmark offset strength corresponding to the steady-state trend channel is queried, and the combined structural offset of the low-frequency trend factor is calculated by combining the market benchmark strength and the structural deviation strength using the following formula:

[0141]

[0142] Among them, δ represents the combined structural offset of the low-frequency trend factor, G adj Indicates the structural deviation strength, G benchmark Indicates the market benchmark deviation strength, max(G adj ,G benchmark ) means selecting the maximum value between the market benchmark strength and the structural deviation strength.

[0143] It should be explained that the steady-state trend constraint parameter is a key quantitative indicator in the steady-state trend channel, which is used to define the characteristics and boundaries of the channel; the net value deviation signal is the prompt information corresponding to the steady-state trend constraint parameter that reflects the deviation of the net value of the investment portfolio from the steady-state trend; the signal sequence is a set of signals corresponding to the net value deviation signal, which is arranged in time or a specific logical order, and is used to present the dynamic changes of the deviation; the structural deviation intensity is a quantitative measure of the severity of the deviation of the investment portfolio structure from the steady-state trend corresponding to the net value deviation signal; the market benchmark deviation intensity is a numerical manifestation of the degree of deviation of the entire market benchmark from the steady-state trend channel corresponding to the steady-state trend channel, which comprehensively reflects the deviation of the overall market from the steady-state trend.

[0144] Furthermore, the steady-state trend constraint parameters in the steady-state trend channel can be calculated through historical data fitting and statistical analysis methods, and the net value deviation signal corresponding to the steady-state trend constraint parameters can be extracted by setting a threshold comparison operation; the signal sequence corresponding to the net value deviation signal can be identified through a time series analysis algorithm, and based on the signal sequence, a specific strength quantification model can be used to calculate the structural deviation intensity corresponding to the net value deviation signal; the market benchmark deviation intensity corresponding to the steady-state trend channel can be queried through the financial data platform interface or the authoritative market database.

[0145] S4. Generate a candidate balancing strategy set corresponding to the portfolio value drift interval based on the risk-return constraint conditions; test the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural deviation; and determine the effective balancing strategy in the candidate balancing strategy set based on the structural deviation tolerance.

[0146] The present invention generates a set of candidate balancing strategies corresponding to the portfolio value drift interval based on the risk-return constraints, thereby providing investors with diverse potential investment strategy options that are tailored to their risk-return demands, helping investors to flexibly position themselves in a complex and changing market environment. Based on the portfolio structure offset, the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set is tested, thereby accurately assessing the ability of each strategy to maintain stable returns in response to changes in the portfolio structure, providing a strong basis for investors to screen out more adaptable and reliable investment strategies. It should be explained that the candidate balancing strategy set is a collection of potential strategies corresponding to the portfolio value drift interval that aim to balance the risk and return of the portfolio and can cause the portfolio value to fluctuate within a given drift interval. These strategies are constructed based on risk-return constraints through an asset allocation model and a strategy generation method. The structural deviation tolerance is a quantitative indicator corresponding to each strategy in the candidate balancing strategy set that measures its ability to maintain a balance between investment objectives and risk and return when the portfolio structure deviates from the steady-state trend channel.

[0147] Specifically, the testing of the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the combined structural deviation includes:

[0148] Performing scene construction processing on the combined structure offset to obtain a structure offset scene;

[0149] Performing a comprehensive simulation process on each strategy in the candidate balancing strategy set and the structural shift scenario, and recording dynamic data of investment indicators during the comprehensive simulation process;

[0150] Identifying a risk-return indicator in the investment indicator dynamic data, and calculating a tolerance score corresponding to the risk-return indicator based on the investment indicator dynamic data;

[0151] Allocating a return indicator weight corresponding to the risk-return indicator, and calculating a comprehensive tolerance score corresponding to each strategy in the candidate balancing strategy set by combining the return indicator weight and the tolerance score;

[0152] Based on the comprehensive tolerance score, the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set is evaluated.

[0153] It should be explained that the structural deviation scenarios are different types of simulation scenarios formed after the portfolio structural deviation is divided according to the dimensions of deviation amplitude, direction and market environment factors, and are used to simulate possible changes in the investment portfolio structure; the investment indicator dynamic data are the dynamic changes in the risk indicators, return indicators and asset allocation ratios of the investment portfolio obtained by real-time recording after comprehensive simulation processing of each strategy in the candidate balance strategy set and the structural deviation scenarios, reflecting the operating status of the investment portfolio of each strategy under different structural deviation scenarios; the risk-return indicators are specific indicators in the investment indicator dynamic data used to measure the risk and return of the investment portfolio, such as portfolio standard deviation, value at risk VaR, annualized rate of return, cumulative rate of return, etc.; The tolerance score is a preliminary score corresponding to the risk-return indicator, obtained by analyzing and quantitatively calculating the risk-return indicators of each strategy in different scenarios based on a pre-established evaluation system, taking into account factors such as the ability of the investment portfolio to maintain risk-return targets over a certain period of time. It reflects the strategy's ability to maintain investment portfolio stability under specific scenarios. The return indicator weight is the weight value assigned to the risk-return indicator based on the importance of different risk-return indicators in evaluating the effectiveness of the strategy and the stability of the investment portfolio. The comprehensive tolerance score is a score calculated by combining the return indicator weight and the tolerance score, which comprehensively measures the strategy's overall ability to maintain investment objectives and risk-return balance under different portfolio structure deviations for each strategy in the candidate balanced strategy set.

[0154] Furthermore, the portfolio structure deviation can be subjected to scenario construction processing by dividing the deviation based on dimensions such as deviation amplitude, direction, and market environment factors to obtain a structural deviation scenario. A comprehensive simulation can be performed on each strategy in the candidate balancing strategy set and the structural deviation scenario using professional investment analysis software or a customized algorithm program, and dynamic investment indicator data during the comprehensive simulation can be recorded by a real-time monitoring system. Risk-return indicators in the dynamic investment indicator data can be identified using pre-set indicator identification rules. Based on the dynamic investment indicator data, a tolerance score corresponding to the risk-return indicator can be calculated using a pre-established evaluation system, taking into account factors such as the ability of the investment portfolio to maintain risk-return targets. Return indicator weights corresponding to the risk-return indicator can be assigned through expert evaluation combined with statistical analysis of historical data. Combining the return indicator weights and the tolerance score, a comprehensive tolerance score corresponding to each strategy in the candidate balancing strategy set can be calculated using mathematical calculation methods such as weighted average. Based on the comprehensive tolerance score, the structural deviation tolerance of each strategy in the candidate balancing strategy set can be evaluated. A higher comprehensive tolerance score indicates a stronger structural deviation tolerance for the strategy, and a better risk-return balance can be maintained when the portfolio structure deviates.

[0155] By determining an effective balancing strategy within the candidate balancing strategy set based on the structural shift tolerance, the present invention can help investors accurately screen strategies that remain robust despite portfolio structural changes, significantly improving the scientific nature and reliability of investment decisions and ensuring the achievement of investment objectives. It should be noted that the effective balancing strategy is a strategy selected from the candidate balancing strategy set based on a comparison of the structural shift tolerance with a preset tolerance threshold. Furthermore, based on a comparison of the structural shift tolerance with a preset tolerance threshold, the effective balancing strategy within the candidate balancing strategy set is determined when the structural shift tolerance exceeds the preset tolerance threshold. The preset tolerance threshold is a pre-set value that can be adjusted appropriately based on reference to the tolerance threshold settings of similar investment portfolios or related financial products in the same industry, taking into account individual investment objectives and characteristics. For example, in a specific investment sector, if the industry generally sets a threshold for a certain risk indicator at 15%, then when constructing a similar investment portfolio, this threshold can be used as a reference to determine whether to use the same threshold or to adjust it upward or downward, based on one's own investment strategy and risk control requirements.

[0156] S5. Combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, construct a spatial risk warning mechanism for the target investment space; based on the spatial risk warning mechanism, perform strategy correction on the effective balance strategy to obtain a corrected balance strategy; and convert the corrected balance strategy into a fund investment recommendation for the target investment space.

[0157] The present invention constructs a spatial risk warning mechanism for the target investment space by combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, comprehensively considering the factors related to capital flow and risk, which can significantly improve the accuracy of risk warning, and based on the spatial risk warning mechanism, the effective balance strategy is revised, thereby enhancing the adaptability and robustness of the fund investment strategy. It should be explained that the spatial risk warning mechanism is a key system for monitoring the target investment space risk and triggering strategy revision. Further, the spatial risk warning mechanism for the target investment space is constructed by combining the liquidity characteristic parameters with the risk resonance attenuation coefficient. The specific steps are as follows: First, in-depth analysis of the liquidity characteristic parameters, including the speed of capital inflow and outflow, trading activity, etc., while refining the risk resonance attenuation coefficient, and clarifying its volatility characteristics under different market conditions. Based on this, a multivariate linear regression model is constructed using historical data to explore the quantitative relationship between the liquidity characteristic parameters and the risk resonance attenuation coefficient, and the initial risk warning threshold is set based on this. Then, a real-time data acquisition system is used to obtain the latest liquidity characteristic parameters and risk resonance attenuation coefficient at a frequency of minutes. Once the data deviates from the warning threshold, the risk assessment program will be immediately initiated. Through Monte Carlo simulation, the probability of risk occurrence and the possible impact in the future will be predicted. Finally, a multi-level warning system will be established. Mild risks will be prompted by system pop-up windows, moderate risks will be notified to relevant personnel via SMS or in-site messages, and severe risks will directly trigger emergency plans, including suspending some high-risk investment operations, adjusting asset allocation structures, etc., to ensure that potential risks can be responded to in a timely and accurate manner within the target investment space.

[0158] The present invention makes strategic corrections to the effective balance strategy based on the spatial risk warning mechanism, thereby being able to respond to market changes in a timely manner, flexibly adjust the fund investment portfolio, reduce potential risks, and improve the stability of investment returns. The modified balance strategy is converted into a fund investment recommendation for the target investment space, which can provide investors with clear investment guidance that fits the real-time market conditions and has both risk prevention and control and return potential, helping investors make more informed and reasonable fund investment decisions. It should be explained that the modified balance strategy is an investment strategy that is adjusted and optimized by the effective balance strategy in response to potential market risks and changes based on the spatial risk warning mechanism; the fund investment recommendation is the modified balance strategy converted into content that is applicable to the target investment space in an easy-to-understand and practical manner, providing investors with specific investment directions and operational guidance.

[0159] Compared with the problems described in the background technology, the present invention extracts the associated fund clusters of the target investment space from the pre-constructed cross-space investment database, and analyzes the risk-return constraints corresponding to the target investment space, which can provide insights into the capital distribution and potential risk-return status of the target investment space, thereby laying an important foundation for the subsequent generation of the candidate balance strategy set corresponding to the portfolio value drift interval. Furthermore, the present invention combines the resource distribution matrix and the flow characteristic parameters, and uses the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, which can clearly present the propagation path and mutual influence relationship of resource fluctuations in different spatial scenarios, so as to improve the accuracy of the subsequent construction of the two-way impact model of the target investment space. The present invention constructs a two-way impact model based on the dynamic resonance spectrum and the resource fluctuation transmission network, which can comprehensively consider the interaction of multiple complex factors in the market, and improve the subsequent generation of candidate balance strategies corresponding to the portfolio value drift interval. The accuracy of the balance strategy set is improved. Furthermore, the present invention generates a candidate balance strategy set corresponding to the portfolio value drift interval based on the risk-return constraint conditions, which can provide investors with diversified potential investment strategy options that are in line with their risk-return demands, helping investors to flexibly layout in a complex and changing market environment. Based on the portfolio structure offset, the structural deviation tolerance corresponding to each strategy in the candidate balance strategy set is tested, which can accurately evaluate the ability of each strategy to maintain stable returns when responding to changes in the portfolio structure, and provide a strong basis for investors to screen out more adaptable and reliable investment strategies. Furthermore, the present invention constructs a spatial risk warning mechanism for the target investment space by combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, comprehensively considering the factors related to capital flow and risk, which can significantly improve the accuracy of risk warning, and based on the spatial risk warning mechanism, the effective balance strategy is revised, thereby enhancing the adaptability and robustness of the fund investment strategy. Therefore, the intelligent fund investment advice generation method and system provided by the embodiment of the present invention can improve the accuracy of fund investment advice generation.

[0160] Example 2:

[0161] like Figure 2 FIG. 1 is a functional module diagram of the intelligent fund investment advice generating system of the present invention.

[0162] The intelligent fund investment advice generation system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the intelligent fund investment advice generation system may include a trend channel construction module 201, a risk resonance attenuation coefficient calculation module 202, a combination structure offset calculation module 203, an effective balance strategy determination module 204, and an advice generation module 205. A module, also referred to as a unit, is a series of computer program segments that can be executed by an electronic device processor and perform a fixed function. These modules are stored in the electronic device's memory.

[0163] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0164] The trend channel construction module 201 is used to extract the associated fund clusters of the target investment space from the pre-built cross-space investment database, analyze the risk-return constraints corresponding to the target investment space, capture the multidimensional spatial signals of the associated fund clusters in real time, decompose the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and respectively construct a dynamic resonance spectrum and a steady-state trend channel related to the emotional pulse intensity of the target investment space;

[0165] The risk resonance attenuation coefficient calculation module 202 is used to collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient of the user investment combination;

[0166] The combined structure offset calculation module 203 is used to construct a two-way impact model of the target input space based on the dynamic resonance map and the resource fluctuation transmission network, calculate the combined value drift interval of the high-frequency resonance factor based on the two-way impact model, and calculate the combined structure offset of the low-frequency trend factor in combination with the steady-state trend channel;

[0167] The effective balancing strategy determination module 204 is configured to generate a set of candidate balancing strategies corresponding to the portfolio value drift interval based on the risk-return constraint condition, test the structural drift tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural drift, and determine an effective balancing strategy in the candidate balancing strategy set based on the structural drift tolerance;

[0168] The recommendation generation module 205 is used to combine the liquidity characteristic parameters with the risk resonance attenuation coefficient to construct a spatial risk warning mechanism for the target investment space, and based on the spatial risk warning mechanism, to modify the effective balance strategy to obtain a modified balance strategy, and convert the modified balance strategy into a fund investment recommendation for the target investment space.

[0169] In detail, each module in the intelligent fund investment suggestion generating system 200 in the embodiment of the present invention adopts the same method as above when in use. Figure 1 The same technical means are used as the method for generating smart fund investment advice described in , and can produce the same technical effects, so I will not go into details here.

[0170] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for generating intelligent fund investment advice, characterized in that: The method comprises: Extracting the associated fund clusters of the target investment space from a pre-built cross-space investment database, analyzing the risk-return constraints corresponding to the target investment space, capturing the multidimensional spatial signals of the associated fund clusters in real time, decomposing the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and constructing a dynamic resonance spectrum and a steady-state trend channel related to the emotional pulse intensity of the target investment space; Collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate the resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient of the user investment combination; Based on the dynamic resonance graph and the resource fluctuation transmission network, a two-way impact model of the target input space is constructed. Based on the two-way impact model, the combined value drift interval of the high-frequency resonance factor is calculated, and the combined structural offset of the low-frequency trend factor is calculated in combination with the steady-state trend channel; generating a candidate balancing strategy set corresponding to the portfolio value drift interval according to the risk-return constraint condition, testing the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural deviation, and determining an effective balancing strategy in the candidate balancing strategy set based on the structural deviation tolerance; Combining the liquidity characteristic parameters with the risk resonance attenuation coefficient, a spatial risk warning mechanism for the target investment space is constructed. Based on the spatial risk warning mechanism, the effective balance strategy is revised to obtain a revised balance strategy, and the revised balance strategy is converted into a fund investment recommendation for the target investment space.

2. The method for generating intelligent fund investment advice according to claim 1, wherein: The analysis of the risk-return constraints corresponding to the target investment space includes: Collecting spatial ecological panoramic data corresponding to the target input space, and performing data cleaning on the spatial ecological panoramic data to obtain target ecological panoramic data; Performing structured processing on the target ecological panoramic data to obtain structured ecological panoramic data; Extracting eco-labels corresponding to the structural ecological panorama data, and analyzing risk-benefit labels from the eco-labels; Based on the risk-benefit label, extracting risk-benefit characteristics from the structural ecological panorama data; Based on the risk-return characteristics, the risk-return constraints corresponding to the target investment space are analyzed.

3. The method for generating intelligent fund investment advice according to claim 1, wherein: Decomposing the multidimensional spatial signal into high-frequency resonance factors and low-frequency trend factors includes: performing signal noise reduction processing on the multidimensional spatial signal to obtain a noise-reduced multidimensional spatial signal; Analyzing the spatial signal dimension corresponding to the denoised multidimensional spatial signal, and constructing a three-dimensional signal tensor of the denoised multidimensional spatial signal based on the spatial signal dimension; Performing missing value completion processing on the signal three-dimensional tensor to obtain a target three-dimensional tensor; Performing matrix decomposition on the target three-dimensional tensor to obtain a basis matrix and a coefficient matrix; A high-frequency resonance factor and a low-frequency trend factor of the multi-dimensional spatial signal are extracted from the base matrix and the coefficient matrix.

4. The method for generating intelligent fund investment advice according to claim 1, wherein: The method of combining the resource distribution matrix and the flow characteristic parameters and using a trained deep learning time series prediction model to generate a resource fluctuation transmission network under different spatial scenarios in the target input space includes: Using the input layer in the deep learning time series prediction model to adjust the resource distribution matrix and the flow characteristic parameters respectively, to obtain an adjusted resource distribution matrix and an adjusted flow characteristic parameter; Calculating the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters using the self-attention function in the deep learning time series prediction model; Utilizing the time series convolution layer in the deep learning time series prediction model, respectively extracting the time series features corresponding to the adjusted resource distribution matrix and the adjusted flow characteristic parameters to obtain matrix time series features and parameter time series features; Analyzing the timing dependency between the matrix timing features and the parameter timing features using the timing cycle layer in the deep learning timing prediction model; Combining the spatiotemporal coupling coefficient and the temporal dependency, and utilizing a fully connected layer in the deep learning temporal prediction model to perform feature fusion on the matrix temporal features and the parameter temporal features to obtain a comprehensive temporal feature; Based on the comprehensive time series features, the network topology optimization unit in the deep learning time series prediction model is used to generate a resource fluctuation transmission network under different spatial scenarios in the target input space.

5. The method for generating intelligent fund investment advice according to claim 4, wherein: The calculating of the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters by using the self-attention function in the deep learning time series prediction model includes: Projecting the adjusted resource distribution matrix onto the corresponding query space to obtain a resource query vector; Projecting the adjusted flow characteristic parameters onto a corresponding key space to obtain a parameter key vector; Combining the resource query vector and the parameter key vector, the self-attention function is used to calculate the spatiotemporal coupling coefficient between the adjusted resource distribution matrix and the adjusted flow characteristic parameters: The self-attention function calculation formula is as follows: Where A represents the spatiotemporal coupling coefficient between the adjustment resource distribution matrix and the adjustment flow characteristic parameters, B(a) represents the resource query vector corresponding to the a-th matrix in the adjustment resource distribution matrix, D(b) represents the parameter key vector corresponding to the b-th parameter in the adjustment flow characteristic parameters, a represents the sequence number of the adjustment resource distribution matrix, b represents the sequence number of the adjustment flow characteristic parameters, and β represents the scaling factor.

6. The method for generating intelligent fund investment advice according to claim 1, wherein: The calculating of the risk resonance attenuation coefficient of the user investment combination includes: Collecting historical rate of return data of the user's investment portfolio, and constructing a rate of return matrix of the user's investment portfolio based on the historical rate of return data; Calculating the covariance matrix corresponding to the yield matrix, performing dynamic conditional correlation coefficient decomposition on the covariance matrix to obtain a time-varying correlation coefficient matrix; Determining a risk resonance event of the user investment combination based on the time-varying correlation coefficient matrix; Calculating the risk resonance intensity corresponding to the risk resonance event, fitting the attenuation factor corresponding to the risk resonance intensity, and defining the half-life corresponding to the risk resonance intensity; Combining the attenuation factor and the half-life, the risk resonance attenuation coefficient of the user investment portfolio is calculated using the following formula: Among them, F represents the risk resonance attenuation coefficient of the user's investment portfolio, λ represents the attenuation factor, and T half Indicates half-life.

7. The method for generating intelligent fund investment advice according to claim 1, wherein: The bidirectional impact model of the target input space is constructed based on the dynamic resonance spectrum and the resource fluctuation transmission network, including: extracting a dynamic resonance feature corresponding to the dynamic resonance spectrum, and performing quantization processing on the dynamic resonance feature to obtain a dynamic resonance feature vector; Performing topological analysis on the resource fluctuation transmission network to obtain network structure characteristics; Performing parameter conversion on the network structure characteristics to obtain a resource fluctuation network parameter set; Combining the dynamic resonance eigenvector and the resource fluctuation network parameter set, a bidirectional impact model of the target input space is constructed.

8. The method for generating intelligent fund investment advice according to claim 1, wherein: The calculating the combined structural offset of the low-frequency trend factor in combination with the steady-state trend channel includes: Calculating a steady-state trend constraint parameter in the steady-state trend channel, and extracting a net value deviation signal corresponding to the steady-state trend constraint parameter; Identifying a signal sequence corresponding to the net value deviation signal, and calculating a structural deviation intensity corresponding to the net value deviation signal based on the signal sequence; The market benchmark offset strength corresponding to the steady-state trend channel is queried, and the combined structural offset of the low-frequency trend factor is calculated by combining the market benchmark strength and the structural deviation strength using the following formula: Among them, δ represents the combined structural offset of the low-frequency trend factor, G adj Indicates the structural deviation strength, G benchmark Indicates the market benchmark deviation strength, max(G adj ,G benchmark ) means selecting the maximum value between the market benchmark strength and the structural deviation strength.

9. The method for generating intelligent fund investment advice according to claim 1, wherein: The step of testing the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the combined structural deviation includes: Performing scene construction processing on the combined structure offset to obtain a structure offset scene; Performing a comprehensive simulation process on each strategy in the candidate balancing strategy set and the structural shift scenario, and recording dynamic data of investment indicators during the comprehensive simulation process; Identifying a risk-return indicator in the investment indicator dynamic data, and calculating a tolerance score corresponding to the risk-return indicator based on the investment indicator dynamic data; Allocating a return indicator weight corresponding to the risk-return indicator, and calculating a comprehensive tolerance score corresponding to each strategy in the candidate balancing strategy set by combining the return indicator weight and the tolerance score; Based on the comprehensive tolerance score, the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set is evaluated.

10. Intelligent fund investment advice generation system, characterized by: The system comprises: A trend channel construction module is used to extract the associated fund clusters of the target investment space from a pre-built cross-space investment database, analyze the risk-return constraints corresponding to the target investment space, capture the multidimensional spatial signals of the associated fund clusters in real time, decompose the multidimensional spatial signals into high-frequency resonance factors and low-frequency trend factors, and respectively construct a dynamic resonance spectrum and a steady-state trend channel related to the emotional pulse intensity of the target investment space; A risk resonance attenuation coefficient calculation module is used to collect the resource distribution matrix and flow characteristic parameters of the user investment combination in the target investment space, combine the resource distribution matrix and the flow characteristic parameters, use the trained deep learning time series prediction model to generate a resource fluctuation transmission network under different spatial scenarios in the target investment space, and calculate the risk resonance attenuation coefficient of the user investment combination; a combined structure offset calculation module, configured to construct a two-way impact model of the target input space based on the dynamic resonance map and the resource fluctuation transmission network, calculate the combined value drift interval of the high-frequency resonance factor based on the two-way impact model, and calculate the combined structure offset of the low-frequency trend factor in combination with the steady-state trend channel; an effective balancing strategy determination module, configured to generate a set of candidate balancing strategies corresponding to the portfolio value drift interval based on the risk-return constraint condition, test the structural deviation tolerance corresponding to each strategy in the candidate balancing strategy set based on the portfolio structural deviation, and determine an effective balancing strategy in the candidate balancing strategy set based on the structural deviation tolerance; A recommendation generation module is used to combine the liquidity characteristic parameters with the risk resonance attenuation coefficient to construct a spatial risk warning mechanism for the target investment space, and based on the spatial risk warning mechanism, to modify the effective balance strategy to obtain a modified balance strategy, and convert the modified balance strategy into a fund investment recommendation for the target investment space.

Images