Method and system for modeling population rearing conditions of saproxylic insects based on data

Abstract

This invention discloses a method and system for determining the propagation conditions of wood-eating insect populations based on data modeling. Specifically, it relates to the field of computer data modeling and simulation optimization technology. The method addresses the problem that existing data modeling methods for establishing propagation conditions of wood-eating insect populations suffer from drift in model parameter mapping relationships and poor reproducibility of output propagation conditions in subsequent batches due to neglecting latent variables such as cross-interface microbial communities. The method establishes an initial mapping model by acquiring environmental parameters and population response data from multiple historical batches. For the data in the initial stage of the current batch, a residual sequence is calculated, and a non-random structure is determined based on the monotonic change trend of the residual fluctuation amplitude with the substrate fermentation process. When a non-random structure exists, the partial dependence relationship between the residual sequence and each environmental factor is extracted along the drift trajectory of the fermentation process as a characterization of the latent variable effect. After concentration screening, the initial mapping model is corrected, and the corrected mapping model is then used to simulate and optimize the output propagation conditions adapted to the latent variable environment.

Description

Technical Field

[0001] This invention relates to the field of computer data modeling and simulation optimization technology, and more specifically, to a method and system for determining the conditions for the propagation of wood-eating insect populations based on data modeling. Background Technology

[0002] In the research and practice of establishing the conditions for the propagation of wood-eating insect populations using computer data modeling, the common approach is to collect environmental state parameters and population response data during artificial propagation. A predictive model reflecting the relationship between population growth and environmental factors is then constructed using regression analysis, artificial neural networks, or differential equation fitting. Based on this model, optimization algorithms are used to search for combinations of conditions such as temperature, humidity, and substrate ratio that maximize population propagation efficiency, ultimately outputting a set of recommended parameters. However, the modeling input features relied upon by this method are concentrated on continuously monitorable macroscopic physical quantities. The actual efficiency of wood-eating insect population propagation is largely affected by a class of biological factors that run between the food substrate and the insect's gut, are difficult to quantify in real time, and are prone to drift with different substrate batches. These factors are unobserved latent variables in the current data acquisition system. When they change significantly between different propagation batches, the mapping relationship originally learned based on macroscopic parameters undergoes a structural shift, causing the propagation conditions output from the model to fail to achieve the expected results in subsequent batches. The reliability and reproducibility of the established conditions are clearly insufficient.

[0003] Existing methods for establishing the propagation conditions of wood-eating insect populations based on data modeling are limited to macroscopically measurable environmental and matrix parameters, while neglecting the modulating effect of biological factors such as interfacial microbial communities as latent variables on population propagation efficiency. When these unmeasurable factors change between different batches, the parameter mapping relationship learned by the model drifts, resulting in poor reproducibility of the propagation conditions obtained by the model in subsequent actual batches and unreliable output results. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for the propagation conditions of wood-eating insect populations based on data modeling to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: Data modeling-based methods for assessing the conditions for population propagation of wood-eating insects include: S1: Obtain environmental parameters and population response data for multiple propagation batches of wood-eating insects, and establish an initial mapping model of population growth rate relative to environmental parameters; S2: For the current batch whose propagation conditions need to be established, the predicted population response is obtained and the residual sequence is calculated using the initial mapping model based on the environmental parameters and population response data of its initial stage. The continuous time window is divided and it is determined whether the residual sequence has a non-random structure. S3: When a non-random structure exists, extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window, and use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as a latent variable to represent its effect. S4: Extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. S5: Use the effect of effective latent variables to characterize the modified initial mapping model and obtain the modified mapping model; S6: Using the environmental parameters of the current batch to be optimized as the search space, the modified mapping model is used to simulate and optimize the population growth rate, and the combination of environmental parameters that meets the population growth rate condition is output as the propagation condition.

[0006] Furthermore, S1 includes: The environmental parameters and population response data of each batch in multiple propagation batches were obtained at the sampling time points during the propagation cycle. The environmental parameters included temperature, humidity, and substrate ratio, and the population response data included the population size at each sampling time point. For each batch, the instantaneous population growth rate is calculated from the population size at adjacent sampling time points. The instantaneous population growth rate is then aligned with the environmental parameters at the corresponding sampling time points to form a sample set of environmental parameters and population growth rate. Multiple batches of environmental parameters and population growth rate samples are merged, and multiple regression fitting is performed on the merged sample set to obtain a regression function with population growth rate as the response variable and environmental parameters as the explanatory variables. The regression function is used as the initial mapping model.

[0007] Furthermore, S2 includes: Input the environmental parameters of the initial stage of the current batch into the initial mapping model, output the predicted population response, and calculate the difference between the actual population response and the predicted population response in the initial stage as the residual sequence. The initial stage is divided into multiple continuous time windows at equal intervals according to the substrate fermentation process. The variance of the residual sequence within each continuous time window is calculated, and the variance is used as the fluctuation amplitude of the residual sequence within that continuous time window. The fluctuation amplitudes of each consecutive time window are arranged in chronological order to obtain the fluctuation amplitude sequence. The monotonicity of the fluctuation amplitude sequence is tested. If the fluctuation amplitude sequence shows a monotonically increasing or monotonically decreasing trend, it is determined that the residual sequence has a non-random structure.

[0008] Furthermore, when determining the monotonic variation trend of the fluctuation amplitude of the residual sequence within each continuous time window as the substrate fermentation process progresses, if the fluctuation amplitude sequence shows a monotonically increasing trend, it is determined that the residual sequence has a non-random structure generated by the strengthening of the cross-interface microbial community as the fermentation process progresses; if the fluctuation amplitude sequence shows a monotonically decreasing trend, it is determined that the residual sequence has a non-random structure generated by the decay of the cross-interface microbial community as the fermentation process progresses.

[0009] Furthermore, S3 includes: For each continuous time window, after removing the linear effects of other environmental factors from the residual sequence and each environmental factor sequence, the partial dependence between the residual sequence and a single environmental factor is calculated, thus obtaining the partial dependence of the residual sequence on each environmental factor under each continuous time window. The partial dependencies of the residual sequences corresponding to each environmental factor under each continuous time window are arranged in chronological order to form a point sequence in the feature space spanned by the environmental factors. The movement path of the point sequence in the feature space is taken as the drift trajectory, and the drift trajectory is used as a latent variable to represent the effect.

[0010] Furthermore, S4 includes: The trajectory points corresponding to each consecutive time window on the drift trajectory are taken as a sub-representation, resulting in a set of sub-representations with the same number of consecutive time windows; In the feature space spanned by environmental factors, the pairwise distances between each sub-representation in the sub-representation set are calculated, and the mean of all pairwise distances is taken as the concentration of the sub-representation set. The concentration is compared with a preset threshold. If the concentration is higher than the preset threshold, the complete trajectory component corresponding to all sub-representations in the drift trajectory is retained as an effective latent variable action representation.

[0011] Furthermore, when comparing the concentration with the preset threshold, the preset threshold is set according to the shape of the drift trajectory of the partial dependence relationship in the feature space under each continuous time window. The more the shape of the drift trajectory is oriented as a straight line, the higher the preset threshold is; the more the shape of the drift trajectory is randomly distributed, the lower the preset threshold is.

[0012] Furthermore, S5 includes: The effective latent variable effect representation is expanded into a correction sequence according to the order of each continuous time window, and the effective latent variable effect representation component corresponding to each continuous time window in the correction sequence is used as the output correction amount of the initial mapping model within the corresponding continuous time window. The corrected output is obtained by subtracting the output correction amount of the corresponding continuous time window from the output of the initial mapping model in each continuous time window. The corrected outputs of all continuous time windows are then integrated to obtain the corrected mapping model.

[0013] Furthermore, S6 includes: The search space is defined as the allowable adjustment range of environmental parameters in the current batch to be optimized. Extract the drift direction of the partial dependency relationship with continuous time window from the effective latent variable action representation, and map the drift direction to the search space to obtain the priority search direction; Multiple sets of candidate environmental parameter combinations are generated in the search space along the preferred search direction. Each set of candidate environmental parameter combinations is input into the modified mapping model to obtain the simulated population growth rate corresponding to each set of candidate environmental parameter combinations. The candidate environmental parameter combinations that simulate a population growth rate that meets the preset population growth rate conditions are output as propagation conditions.

[0014] On the other hand, the present invention provides a data modeling-based system for the propagation of wood-eating insect populations, comprising: The initial modeling module is used to acquire environmental parameters and population response data from multiple propagation batches of wood-eating insects and establish an initial mapping model of population growth rate relative to environmental parameters. The residual determination module is used to obtain the predicted population response and calculate the residual sequence based on the environmental parameters and population response data of the current batch to be established for propagation conditions, using the initial mapping model to determine the residual sequence, dividing the continuous time window and determining whether the residual sequence has a non-random structure. The latent variable representation module is used to extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window when a non-random structure exists, and to use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as the latent variable representation. The effective filtering module is used to extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. The model correction module is used to correct the initial mapping model by utilizing the effective latent variables to obtain the corrected mapping model. The condition optimization module is used to simulate and optimize the population growth rate using the environmental parameters of the current batch to be optimized as the search space and the modified mapping model. It outputs the combination of environmental parameters that meet the population growth rate conditions as the propagation conditions.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. By extracting the monotonic variation trend of the residual sequence's fluctuation amplitude within each continuous time window, it is determined whether the initial mapping model has a non-random structure in the current batch due to latent variable interference. The drift trajectory of the partial dependence relationship between the residual sequence and each environmental factor under the non-random structure along the substrate fermentation process is used as a latent variable characterization. This characterization method does not require direct measurement of the cross-interface microbial community, but captures the dynamic fingerprint of latent variable modulation from existing environmental parameters and population response data, making the influence of latent variables explicit at the data level. Based on this, the concentration of drift trajectory components within each continuous time window is used to screen out effective latent variable characterizations. Then, the effective latent variable characterizations are converted into output correction values ​​to perform window-by-window compensation and refitting of the initial mapping model, so that the corrected mapping model is embedded with the latent variable modulation mode unique to the current batch. Thus, at the computer modeling level, the structural shift problem of mapping relationship caused by latent variable drift between batches is solved.

[0016] 2. The modified mapping model's predicted output for population growth rate has corrected the systematic bias caused by latent variables. When using this model as a proxy for simulation optimization in the search space, the priority search direction is guided by the drift direction of the partial dependency relationship, making the optimization process conform to the evolution trend of latent variables in the current batch. The generated combination of environmental parameters is closer to the true optimal response region of the current batch's propagation system. From residual structure analysis, latent variable characterization extraction, characterization reliability screening, model adaptive correction to condition simulation optimization, all are completed within the scope of data acquisition and model operation, without relying on physical intervention in the propagation process. The output propagation conditions are batch-specific and reproducible, improving the reliability of the technical path of establishing wood-eating insect population propagation conditions based on computer modeling in dealing with raw material batch differences. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for the population propagation of wood-eating insects based on data modeling according to the present invention; Figure 2 This is a schematic diagram of the structure of the data modeling-based system for the propagation of wood-eating insect populations according to the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention presents a data modeling-based method for determining the conditions for the propagation of wood-eating insect populations, including: S1: Obtain environmental parameters and population response data for multiple propagation batches of wood-eating insects, and establish an initial mapping model of population growth rate relative to environmental parameters; S2: For the current batch whose propagation conditions need to be established, the predicted population response is obtained and the residual sequence is calculated using the initial mapping model based on the environmental parameters and population response data of its initial stage. The continuous time window is divided and it is determined whether the residual sequence has a non-random structure. S3: When a non-random structure exists, extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window, and use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as a latent variable to represent its effect. S4: Extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. S5: Use the effect of effective latent variables to characterize the modified initial mapping model and obtain the modified mapping model; S6: Using the environmental parameters of the current batch to be optimized as the search space, the modified mapping model is used to simulate and optimize the population growth rate, and the combination of environmental parameters that meets the population growth rate condition is output as the propagation condition.

[0020] S1: Obtain environmental parameters and population response data for multiple propagation batches of wood-eating insects, establish an initial mapping model of population growth rate relative to environmental parameters, and implement it as follows: Environmental parameters and population response data for each batch within multiple propagation batches were recorded at sampling time points during the propagation period. Environmental parameters included temperature, humidity, and substrate ratio, while population response data included the population size at each sampling time point. The propagation of white-spotted flower beetle larvae on corn stalk substrate was used as an example. For each batch, a propagation period was pre-set, representing the total time from larval inoculation to larval harvest. Within the propagation period, sampling time points were set at preset fixed time intervals, such as 24 hours. Environmental parameters and population response data were recorded every 24 hours from the inoculation time. Temperature was read using a temperature sensor 1 cm below the substrate surface in the rearing container, and humidity was read using a humidity sensor 2 cm above the substrate surface. The substrate ratio was recorded as a percentage by mass, representing the proportion of corn stalks to added materials, for example, 80% corn stalks and 20% wheat bran, or 70% corn stalks and 30% wheat bran. The specific proportions were determined based on the actual weight of materials used in each batch. The population response data is recorded as follows: at each sampling time point, all larvae are removed from the rearing container and counted. The count result is used as the population size at that sampling time point. After counting, the larvae are returned to the rearing container for continued rearing. Thus, each batch generates a record of environmental parameters and population response data arranged according to the sampling time points within the propagation cycle.

[0021] For each batch, the instantaneous population growth rate is calculated from the population size at adjacent sampling time points. This instantaneous growth rate is then aligned with the environmental parameters at the corresponding sampling time points to form a sample set of environmental parameters and population growth rates. For a single batch, the population size sequence arranged by sampling time points is extracted. The population size at each of two adjacent sampling time points is then taken. The population size at the later sampling time point is subtracted from the population size at the earlier sampling time point. The difference is divided by the population size at the earlier sampling time point, and then by the time interval between the two adjacent sampling time points to obtain the instantaneous population growth rate between these two sampling time points. The calculated instantaneous population growth rate is then mapped to the environmental parameters at the later sampling time point. That is, for any pair of adjacent sampling time points, the instantaneous population growth rate is combined with the temperature, humidity, and substrate ratio recorded at the later sampling time point to form a sample record. This process is repeated for all pairs of adjacent sampling time points within the batch to obtain the sample set of environmental parameters and population growth rates for that batch. This alignment method directly correlates the instantaneous population growth rate with the environmental conditions that generated the instantaneous population growth rate in time, reflecting the actual instantaneous changes in the population under the influence of current environmental parameters.

[0022] Multiple batches of environmental parameters and population growth rate sample sets are merged. A multiple regression model is then performed on the merged sample set to obtain a regression function with population growth rate as the response variable and environmental parameters as explanatory variables. This regression function serves as the initial mapping model. The environmental parameters and population growth rate sample sets obtained independently from each batch are vertically merged, meaning all batch sample records are placed into the same data table to form a merged sample set. When performing multiple regression on the merged sample set, the population growth rate is used as the response variable, and temperature, humidity, and substrate ratio are used as explanatory variables. A multiple multinomial regression model is constructed, including constant terms, linear terms, quadratic terms, and pairwise interaction terms. The regression model takes the form: Population growth rate is calculated as: R = β0 + β1×T + β2×H + β3×S + β4×T 2 +β5×H 2 +β6×S 2 The regression model is defined as follows: R = +β7×T×H+β8×T×S+β9×H×S; where R represents the population growth rate, T represents temperature, H represents humidity, S represents substrate ratio, β0 represents the constant term coefficient, β1 to β3 represent the linear coefficients of temperature, humidity, and substrate ratio, respectively, β4 to β6 represent the quadratic coefficients of temperature, humidity, and substrate ratio, respectively, and β7 to β9 represent the interaction coefficients of temperature and humidity, temperature and substrate ratio, and humidity and substrate ratio, respectively. The least squares method is used to estimate the coefficients of each term in the regression model, minimizing the sum of squared residuals between the actual population growth rate of all sample records in the merged sample set and the predicted population growth rate calculated by substituting the corresponding environmental parameters into the regression model. The values ​​of β0 to β9 are then obtained by solving for these values. Substituting these coefficient values ​​into the regression model yields the regression function. The regression function is used as the initial mapping model, which establishes a quantitative mapping relationship between environmental parameters and population growth rate. By inputting any set of values ​​for temperature, humidity and substrate ratio, the initial mapping model calculates and outputs the corresponding predicted population growth rate value through the regression function.

[0023] S2: For the current batch whose propagation conditions need to be established, based on its initial environmental parameters and population response data, the predicted population response is obtained using the initial mapping model, and the residual sequence is calculated. Continuous time windows are divided, and it is determined whether the residual sequence has a non-random structure. The implementation is as follows: After multiple consecutive propagation batches, environmental parameters and population response data from different batches were accumulated. Using this data, an initial mapping model capable of predicting population growth rate based on environmental parameters was established in S1. When it is necessary to establish propagation conditions for an upcoming or newly started propagation batch, this batch is designated as the current batch, and data collection and model validation are performed on its initial stage. The initial stage of the current batch refers to a period of pre-propagation starting from the inoculation time. The duration of this initial stage does not cover the entire propagation cycle, but only a few sampling intervals in the early stages, such as from the inoculation time to the first 72 hours of the propagation cycle. During the initial stage of the current batch, the sampling method consistent with S1 is used. Temperature, humidity, and substrate ratio are recorded at each sampling time point as the environmental parameter sequence for the initial stage of the current batch, and the population size at each sampling time point is recorded as the population response sequence for the initial stage of the current batch.

[0024] The temperature, humidity, and substrate ratio at each sampling time point in the environmental parameter sequence collected during the initial stage of the current batch are sequentially input into the initial mapping model established in S1. The initial mapping model outputs a predicted population growth rate for each set of input environmental parameters. The predicted population growth rates for each sampling time point during the initial stage of the current batch are arranged chronologically to form a predicted population response sequence. Simultaneously, the actual population count sequence obtained from the aquaculture containers during the initial stage of the current batch is used as the actual population response sequence. The population growth rate at each sampling time point in the actual population response sequence is subtracted from the predicted population growth rate at the corresponding sampling time point in the predicted population response sequence. The resulting difference sequence is the residual sequence.

[0025] The residual sequence may contain both irregular fluctuations caused by random measurement errors and directional systematic biases resulting from changes in the cross-interfacial microbial community during fermentation. To distinguish between these two possibilities, the substrate fermentation process is used as a time reference to analyze the structure of the residual sequence. The substrate fermentation process refers to the degree of directional succession of the microbial community in the substrate caused by the feeding and excretion activities of wood-eating insects from the inoculation time, and the substrate fermentation process deepens continuously over time. The initial stage of the current batch is divided into multiple consecutive time windows at equal intervals according to the substrate fermentation process. The division method is as follows: the total duration of the initial stage of the current batch is divided by the preset number of time windows to obtain the time span of each consecutive time window. For example, if the 72 hours of the initial stage of the current batch is divided into 3 consecutive time windows at equal intervals, then the time span of each consecutive time window is 24 hours, or if it is divided into 4 consecutive time windows at equal intervals, then the time span of each consecutive time window is 18 hours. The number of time windows is determined based on the total number of sampling points in the initial stage of the current batch, to ensure that the number of sampling points included in each consecutive time window is sufficient to support variance calculation.

[0026] Within each continuous time window, all residual values ​​contained within the window are extracted, and the variance of the residual sequence within the continuous time window is calculated. The variance is calculated as follows: first, the mean of all residual values ​​within the continuous time window is calculated; then, the square of the difference between each residual value and the mean is calculated; the sum of squares is averaged, and the resulting average is the variance of the residual sequence within the continuous time window. The variance of the residual sequence within the continuous time window is used as the fluctuation range of the residual sequence within the continuous time window.

[0027] Arrange the continuous time windows in chronological order, and sequentially extract the fluctuation amplitude of the residual sequence corresponding to each continuous time window to obtain a fluctuation amplitude sequence that corresponds one-to-one with the sequence of continuous time windows. Perform a monotonicity test on the fluctuation amplitude sequence. The method for the monotonicity test is as follows: calculate the difference value between the fluctuation amplitudes of adjacent continuous time windows in the fluctuation amplitude sequence, that is, subtract the fluctuation amplitude of the previous continuous time window from the fluctuation amplitude of the later continuous time window to obtain the difference value sequence; traverse the difference value sequence. If all difference values ​​in the difference value sequence are positive, the fluctuation amplitude sequence shows a monotonically increasing trend; if all difference values ​​in the difference value sequence are negative, the fluctuation amplitude sequence shows a monotonically decreasing trend; if the difference value sequence contains both positive and negative values, the fluctuation amplitude sequence does not show a monotonically changing trend.

[0028] If the fluctuation amplitude sequence shows a monotonically increasing trend, it indicates that the fluctuation amplitude of the residual sequence continuously increases with the deepening of substrate fermentation. The prediction bias of the initial mapping model shows a directional dependence of the substrate fermentation process, indicating that the residual sequence has a non-random structure. Furthermore, this non-random structure is attributed to the strengthening of the cross-interfacial microbial community as fermentation progresses. If the fluctuation amplitude sequence shows a monotonically decreasing trend, it indicates that the fluctuation amplitude of the residual sequence continuously narrows with the deepening of substrate fermentation. Similarly, the prediction bias of the initial mapping model shows a directional dependence of the substrate fermentation process, indicating that the residual sequence has a non-random structure. Furthermore, this non-random structure is attributed to the decay of the cross-interfacial microbial community as fermentation progresses. If the fluctuation amplitude sequence shows neither a monotonically increasing nor a monotonically decreasing trend, the fluctuation amplitude change of the residual sequence lacks directionality. The deviation in the residual sequence originates from random measurement errors rather than the systematic modulation effect of the cross-interfacial microbial community, indicating that the residual sequence does not have a non-random structure.

[0029] S3: When a non-random structure exists, the partial dependencies between the residual sequences and each environmental factor are extracted according to each continuous time window. The drift trajectory of the partial dependencies under each continuous time window in the feature space is used as a latent variable representation, as follows: After determining the existence of a non-random structure in the residual sequence in S2, it indicates that the prediction bias of the initial mapping model in the current batch contains a systematic component generated by the changes in the cross-interface microbial community as the substrate fermentation process progresses. To make this systematic component explicit under conditions where the composition and activity of the cross-interface microbial community cannot be directly measured, it is necessary to extract latent variable representations reflecting the modulating effect of the cross-interface microbial community from the relationship between the residual sequence and environmental factors. The modulation of the population growth rate by the cross-interface microbial community is not independent of environmental factors, but rather indirectly affects the population growth rate through synergistic or antagonistic relationships with temperature, humidity, and substrate ratio. When the cross-interface microbial community drifts between different batches, the way environmental factors affect the population growth rate also changes, and this change is reflected in the dynamic changes in the dependency relationship between the residual sequence and environmental factors. Therefore, by extracting the partial dependency relationship between the residual sequence and each environmental factor window-by-window within each continuous time window, and then observing the trajectory of the partial dependency relationship between each continuous time window, it is possible to capture the dynamic pattern of the modulating effect of the cross-interface microbial community as a latent variable on the population growth rate.

[0030] For each continuous time window, partial dependencies between the residual sequence and each environmental factor are extracted. Partial dependency refers to the net correlation remaining between the residual sequence and a single environmental factor after removing the linear effects of other environmental factors on the residual sequence. The specific steps for extracting partial dependencies are as follows: For each continuous time window obtained in S2, the temperature, humidity, and matrix ratio sequences falling within the continuous time window are extracted from the environmental parameter sequences of the initial stage of the current batch. Simultaneously, residual sequence segments falling within the same continuous time window are extracted from the residual sequences calculated in S2. Taking the calculation of the partial dependence between the residual sequence and temperature within a continuous time window as an example, firstly, the residual sequence segment is used as the dependent variable, and the humidity sequence and matrix ratio sequence are used as independent variables. A multiple linear regression is performed to obtain the remaining part of the residual sequence segment after being linearly explained by the humidity sequence and matrix ratio sequence, which is denoted as the first residual sequence. Then, the temperature sequence is used as the dependent variable, and the humidity sequence and matrix ratio sequence are used as independent variables. A multiple linear regression is performed to obtain the remaining part of the temperature sequence after being linearly explained by the humidity sequence and matrix ratio sequence, which is denoted as the second residual sequence. The Pearson correlation coefficient between the first residual sequence and the second residual sequence is calculated. The Pearson correlation coefficient is the partial dependence between the residual sequence and temperature within the continuous time window. Similarly, when calculating the partial dependence between the residual sequence and humidity within a continuous time window, the residual sequence segment is first used as the dependent variable, and the temperature and matrix ratio sequences are used as independent variables to perform multiple linear regression to obtain the residual sequence. Then, the humidity sequence is used as the dependent variable, and the temperature and matrix ratio sequences are used as independent variables to perform multiple linear regression to obtain the residual sequence. The Pearson correlation coefficient between the two residual sequences is the partial dependence between the residual sequence and humidity within the continuous time window. When calculating the partial dependence between the residual sequence and matrix ratio within a continuous time window, the residual sequence segment is first used as the dependent variable, and the temperature and humidity sequences are used as independent variables to perform multiple linear regression to obtain the residual sequence. Then, the matrix ratio sequence is used as the dependent variable, and the temperature and humidity sequences are used as independent variables to perform multiple linear regression to obtain the residual sequence. The Pearson correlation coefficient between the two residual sequences is the partial dependence between the residual sequence and matrix ratio within the continuous time window. Iterate through all the continuous time windows obtained in S2, calculate a set of partial dependence values ​​of temperature, humidity and matrix ratio corresponding to the residual sequence for each continuous time window, and obtain the partial dependence of each environmental factor corresponding to the residual sequence under each continuous time window.

[0031] The partial dependencies of the residual sequences corresponding to each environmental factor under each continuous time window are arranged in chronological order according to the continuous time window, forming a point sequence in the feature space spanned by the environmental factors. The feature space spanned by the environmental factors is a three-dimensional space with the partial dependency of the residual sequence corresponding to temperature as the first coordinate axis, the partial dependency of the residual sequence corresponding to humidity as the second coordinate axis, and the partial dependency of the residual sequence corresponding to matrix ratio as the third coordinate axis. These three coordinate axes constitute the feature space. For each continuous time window, the partial dependency value of the residual sequence corresponding to temperature is used as the first coordinate value, the partial dependency value of the residual sequence corresponding to humidity as the second coordinate value, and the partial dependency value of the residual sequence corresponding to matrix ratio as the third coordinate value. These three coordinate values ​​together determine a point in the feature space, which is called a trajectory point in the feature space. A trajectory point is generated for each consecutive time window according to their chronological order. Connecting these trajectory points sequentially creates a path in the feature space, pointing from the trajectory point of the earliest consecutive time window to the trajectory point of the latest consecutive time window. This path represents the drift trajectory of the partial dependency relationship within each consecutive time window in the feature space. The drift trajectory visually depicts the directional change in the modulatory effect of the cross-interface microbial community on the population growth rate along the environmental factor dimension as the substrate fermentation process progresses. Using the drift trajectory as a latent variable representation—a path with a definite start and end point in the feature space—records how the cross-interface microbial community gradually alters the net association pattern between the residual sequence and various environmental factors throughout the initial stage of the current batch.

[0032] S4: Extract sub-representations from the drift trajectory within each consecutive time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable representations, as follows: Not all trajectory points on the drift trajectory necessarily reflect the true modulation effect of the cross-interface microbial community. The initial sampling process may introduce some random noise, or the partial dependency calculation of the residual sequence within a certain continuous time window may be interfered with by abnormal sampling values, causing some trajectory points on the drift trajectory to deviate from the true direction of the cross-interface microbial community modulation. If the entire drift trajectory containing noise interference is used to subsequently revise the initial mapping model, the revision effect will be contaminated by noise, resulting in a decrease in the prediction accuracy of the revised mapping model. Therefore, it is necessary to evaluate the consistency of each trajectory point on the drift trajectory in the feature space spanned by environmental factors, and select the drift trajectory components corresponding to the trajectory points that are concentrated and close to each other in the feature space, retaining them as effective latent variable effects, while removing the scattered and far-away trajectory points from the latent variable effects representation, to ensure that the signal on which the model revision is based comes from the true modulation effect of the cross-interface microbial community rather than random noise.

[0033] Each trajectory point corresponding to a consecutive time window on the drift trajectory is treated as a sub-representation, resulting in a set of sub-representations equal to the number of consecutive time windows. A sub-representation is a local representation of the drift trajectory within a single consecutive time window. The coordinate values ​​of a sub-representation are equal to the partial dependence values ​​of the residual sequence corresponding to temperature, humidity, and matrix ratio within the consecutive time window. Each sub-representation in the feature space spanned by environmental factors is represented as a point with three coordinate values. All trajectory points from the first to the last consecutive time window on the drift trajectory are extracted one by one to form a set of sub-representations. The number of sub-representations in the set is equal to the number of consecutive time windows divided in S2.

[0034] In a feature space spanned by environmental factors, the pairwise distances between each sub-representation in the sub-representation set are calculated, and the mean of all pairwise distances is taken as the centrality of the sub-representation set. The pairwise distance between sub-representations refers to the length of the line segment connecting corresponding points of two sub-representations in the feature space. The distance between any two sub-representations is calculated as follows: Take two sub-representations from the sub-representation set, obtain the three coordinate values ​​of the first sub-representation and the second sub-representation in the feature space. Subtract the coordinate values ​​of the two sub-representations on the first coordinate axis and take the square; subtract the coordinate values ​​of the two sub-representations on the second coordinate axis and take the square; subtract the coordinate values ​​of the two sub-representations on the third coordinate axis and take the square. Sum the three squares and take the square root; the square root value is the pairwise distance between the two sub-representations. Iterate through all unique pairs of sub-representations in the sub-representation set, calculate the pairwise distance between each pair, and obtain several pairwise distance values. Sum all pairwise distance values ​​and divide by the total number of pairwise distance values; the average value is the centrality of the sub-representation set. The lower the concentration, the more closely the sub-representations in the sub-representation set are distributed in the feature space, the more consistent the partial dependencies of each continuous time window are, and the more compact the shape of the drift trajectory. The higher the concentration, the more dispersed the sub-representations in the sub-representation set are distributed in the feature space, the greater the difference in the partial dependencies of each continuous time window are, and the looser the shape of the drift trajectory.

[0035] The concentration is compared with a preset threshold. If the concentration is less than the preset threshold, the complete trajectory components of all corresponding sub-representations in the drift trajectory are retained as valid latent variable effects. The preset threshold is a boundary value used to determine whether the concentration of the sub-representation set meets the reliability requirements. The preset threshold is set according to the shape of the drift trajectory of the partial dependency relationship in the feature space under each continuous time window. The more the shape of the drift trajectory tends to be a directional straight line, the more it indicates that the sub-representations in the sub-representation set are arranged in an orderly manner along a single direction, and the modulation effect of the cross-interface microbial community has a clear direction and consistency. The preset threshold is set to a lower value. The more the shape of the drift trajectory tends to be randomly scattered, the more it indicates that the sub-representations in the sub-representation set are randomly distributed in the feature space, and the modulation effect of the cross-interface microbial community is unclear. The preset threshold is set to a higher value. The preset threshold is set as follows: First, the drift trajectory is linearly fitted to obtain a fitted straight line. The vertical distance from each trajectory point on the drift trajectory to the fitted straight line is calculated, and the mean of the vertical distances is used as the trajectory deviation. The trajectory deviation is added to a constant 1, and the reciprocal is taken. This reciprocal is then multiplied by a preset benchmark value to obtain the preset threshold. The preset benchmark value is determined based on the statistical distribution of the mean of pairwise distances in multiple historical batches of sub-representation sets. For example, the upper quartile of the mean of pairwise distances in multiple historical batches of sub-representation sets is used as the preset benchmark value. The larger the trajectory deviation, the more random the drift trajectory tends to be, and the smaller the factor obtained by adding the trajectory deviation to a constant 1 and taking the reciprocal. The preset threshold obtained by multiplying it by the preset benchmark value is still at a relatively high level. Conversely, the smaller the trajectory deviation, the more directional the drift trajectory tends to be, and the larger the factor obtained by adding the trajectory deviation to a constant 1 and taking the reciprocal. The preset threshold obtained by multiplying it by the preset benchmark value is lowered to a more stringent level. If the concentration is less than a preset threshold, it indicates that the distribution density of the sub-representation set meets the reliability standard required by the drift trajectory morphology. The complete trajectory components corresponding to all sub-representations in the drift trajectory are retained as effective latent variable representations for subsequent correction of the initial mapping model. If the concentration is not less than the preset threshold, it indicates that the distribution of the sub-representation set is too dispersed, posing a risk of misjudging random interference from individual continuous time windows as a systematic modulation signal. Therefore, the complete trajectory components of the drift trajectory are not retained as effective latent variable representations.

[0036] S5: Utilize the effective latent variables to characterize the modified initial mapping model, resulting in the modified mapping model, implemented as follows: The effective latent variable representation retained after concentration screening in S4 is a drift trajectory component with a clear direction and shape in a feature space spanned by environmental factors. This effective latent variable representation records the modulation pattern of the cross-interfacial microbial community on the partial dependence relationship between the residual sequence and various environmental factors within each consecutive time window of the current batch. The initial mapping model in S1 is established through multiple regression fitting based on environmental parameters and population growth rate sample sets from multiple historical batches. The population growth rate response to environmental parameters described by the initial mapping model does not include the specific modulation effect of the cross-interfacial microbial community in the current batch. When S2 determines that the residual sequence has a non-random structure, and after extraction in S3 and screening in S4, it is confirmed that the non-random structure of the residual sequence originates from the systematic modulation of the cross-interfacial microbial community, the predicted output of the initial mapping model within each consecutive time window of the current batch will have a systematic deviation from the actual population growth rate within the consecutive time windows due to the modulation effect of the cross-interfacial microbial community. The initial mapping model is corrected by using the effective latent variable action representation. Specifically, the prediction bias of the initial mapping model is compensated window by window by the modulation mode of each continuous time window recorded in the effective latent variable action representation, so that the corrected mapping model can be adapted to the cross-interface microbial community environment of the current batch.

[0037] The effective latent variable impact representation is expanded into a correction sequence according to the order of each continuous time window. The correction sequence is a one-dimensional data sequence composed of the effective latent variable impact representation components corresponding to each continuous time window arranged in chronological order. An effective latent variable impact representation component refers to the local representation quantity corresponding to a single continuous time window in the effective latent variable impact representation. For each continuous time window, the effective latent variable impact representation component is a point with three coordinate values. These three coordinate values ​​correspond to the partial dependence values ​​between the residual sequence and temperature, humidity, and matrix ratio within the continuous time window, respectively. The effective latent variable impact representation component composed of these three coordinate values ​​is converted into the output correction quantity of the initial mapping model within the corresponding continuous time window. The output correction quantity is a scalar value.

[0038] Calculate the output correction: Δ = wT × pT + wH × pH + wS × pS; where Δ represents the output correction for the corresponding continuous time window, pT represents the partial dependence of the residual sequence on temperature within the corresponding continuous time window, pH represents the partial dependence of the residual sequence on humidity within the corresponding continuous time window, pS represents the partial dependence of the residual sequence on matrix ratio within the corresponding continuous time window, wT represents the influence ratio of temperature within the corresponding continuous time window, wH represents the influence ratio of humidity within the corresponding continuous time window, and wS represents the influence ratio of matrix ratio within the corresponding continuous time window.

[0039] The proportion of the effect of temperature is calculated as: wT = σ 2 T÷(σ 2 T+σ 2 H+σ 2 S); The calculation of the proportion of the influence of humidity is: wH=σ 2 H÷(σ 2 T+σ 2 H+σ 2 S); The influence ratio of the matrix ratio is calculated as follows: wS=σ 2 S÷(σ 2 T+σ 2 H+σ 2 S); where σ 2 T represents the variance of the temperature within the corresponding continuous time window, σ 2 H represents the variance of humidity within the corresponding continuous time window, σ 2 S represents the variance of the matrix ratio within the corresponding continuous time window. For example, if the variance of temperature is 0.5, the variance of humidity is 0.3, and the variance of matrix ratio is 0.2 within a certain continuous time window, and the total variance of temperature, humidity, and matrix ratio is 1.0, then the influence ratio of temperature is 0.5, the influence ratio of humidity is 0.3, and the influence ratio of matrix ratio is 0.2. Multiplying the three partial dependency values ​​by their respective influence ratios and summing the results gives the output correction for the corresponding continuous time window. Arranging the output corrections for each continuous time window sequentially according to their chronological order forms a correction sequence, where each element in the correction sequence corresponds one-to-one with a continuous time window.

[0040] The corrected output is obtained by subtracting the corresponding output correction amount from the output of the initial mapping model within each continuous time window. The corrected outputs of all continuous time windows are then integrated to obtain the corrected mapping model. For each continuous time window divided in S2, the population growth rate prediction sequence calculated by the initial mapping model based on the input environmental parameter sequence is extracted. Each predicted population growth rate in the sequence corresponds to a sampling time point within that continuous time window. The output correction amount corresponding to that continuous time window is subtracted from all predicted population growth rates in the sequence to obtain the corrected population growth rate sequence. This corrected population growth rate sequence is the estimated population growth rate within that continuous time window after latent variable compensation. All continuous time windows are traversed, and the corrected population growth rate sequences of each continuous time window are concatenated end-to-end according to the time order of the continuous time windows. This concatenation yields a corrected population growth rate sequence covering the entire initial stage of the current batch. The corrected population growth rate sequence for the entire initial stage of the current batch is then realigned with the corresponding initial stage environmental parameter sequence to form a corrected environmental parameter and population growth rate sample set. The corrected environmental parameters and population growth rate sample set were refitted using multiple regression. The regression model used during fitting was consistent with the one used in establishing the initial mapping model in S1. The regression model included a constant term, a linear term for temperature, a linear term for humidity, a linear term for substrate ratio, a squared term for temperature, a squared term for humidity, a squared term for substrate ratio, an interaction term for temperature and humidity, an interaction term for temperature and substrate ratio, and an interaction term for humidity and substrate ratio. The coefficients of each term in the regression model were solved using the least squares method. The obtained coefficients were substituted into the form of the regression model to obtain the corrected mapping model. Based on the functional form of the initial mapping model, the corrected mapping model incorporates the modulation pattern of the current batch of cross-interfacial microbial communities through refitting. The population growth rate prediction value output by the corrected mapping model for the input environmental parameters has compensated for the systematic bias caused by the cross-interfacial microbial community.

[0041] S6: Using the environmental parameters of the current batch to be optimized as the search space, the modified mapping model is used to simulate and optimize the population growth rate. The combination of environmental parameters that meets the population growth rate condition is output as the propagation condition. The implementation is as follows: The modified mapping model incorporates the modulation pattern of the current batch of cross-interfacial microbial communities, enabling it to output predicted population growth rates after latent variable compensation for input environmental parameters. The application goal of the modified mapping model is to determine a set of environmental parameters for the optimization phase of the current batch, ensuring that the population growth rate meets preset conditions. This set of environmental parameters is then output as propagation conditions to guide subsequent propagation operations in the current batch. The optimization phase of the current batch refers to the propagation phase after the initial phase. During this phase, the environmental parameters are not yet determined, requiring simulation optimization based on the modified mapping model to obtain the optimal environmental parameter settings.

[0042] The search space is defined by the allowable adjustment range of the environmental parameters for the current batch in the optimization stage. These parameters include temperature, humidity, and substrate ratio. Each parameter has an allowable adjustment range, determined by the controllable range of the propagation facility and the tolerance boundary of the wood-eating insects. For example, the allowable temperature range is 20°C to 35°C, the humidity range is 60% to 90%, and the allowable substrate ratio range is a continuous range from the lower to the upper limit of the mass percentage of auxiliary materials in the substrate ratio (e.g., 10% to 40%). The Cartesian product of these three allowable ranges forms a three-dimensional continuous region, which is the search space. Each point in the search space corresponds to a set of combinations of temperature, humidity, and substrate ratio values.

[0043] The drift direction of partial dependencies with continuous time windows is extracted from the effective latent variable representation. This drift direction is then mapped to the search space to obtain the preferred search direction. The effective latent variable representation is the drift trajectory component retained in S4. The drift trajectory is formed by connecting the trajectory points of each continuous time window in the feature space spanned by environmental factors. The drift trajectory has a directionality from the trajectory point corresponding to the earliest continuous time window to the trajectory point corresponding to the latest continuous time window. The drift direction is extracted by taking the trajectory points corresponding to the first and last continuous time windows on the drift trajectory, and calculating the vector from the trajectory point corresponding to the first continuous time window to the trajectory point corresponding to the last continuous time window. This vector represents the drift direction of the partial dependencies with continuous time windows. The drift direction reflects the evolution trend of the modulatory effect of the cross-interface microbial community in the entire initial stage of the current batch, i.e., in which direction the partial dependencies between the residual sequence and temperature, humidity, and substrate ratio change as the substrate fermentation process progresses. The drift direction is mapped to the search space as follows: the components of the drift direction on the three coordinate axes correspond to the evolution directions of the partial dependencies between the residual sequence and temperature, humidity, and matrix ratio, respectively. These three components are used as the search direction components for the temperature, humidity, and matrix ratio dimensions in the search space. The vector synthesized from these three search direction components is the preferred search direction in the search space. The preferred search direction points to the direction in which the population growth rate response improves the most under the modulation of the cross-interfacial microbial community in the current batch.

[0044] Multiple candidate environmental parameter combinations are generated along the preferred search direction in the search space. Each candidate environmental parameter combination is input into the modified mapping model to obtain the simulated population growth rate corresponding to each combination. The method for generating multiple candidate environmental parameter combinations is as follows: the mean value of the environmental parameters within the last continuous time window of the initial stage of the current batch is used as the search starting point. Starting from the search starting point, the search proceeds along the preferred search direction with a preset step size. The preset step size is determined based on the allowable adjustment range width of each dimension in the search space. For example, the preset step size for the temperature dimension is 5% of the allowable temperature adjustment range width, the preset step size for the humidity dimension is 5% of the allowable humidity adjustment range width, and the preset step size for the matrix ratio dimension is 5% of the allowable matrix ratio adjustment range width. Each preset step size generates one candidate environmental parameter combination. During the generation of candidate environmental parameter combinations, boundary checks are performed on each combination. If any environmental parameter in a candidate combination exceeds the boundary of its corresponding allowable adjustment range, the environmental parameter exceeding the boundary is adjusted to the corresponding boundary value before being included as a candidate environmental parameter combination. Multiple sets of candidate environmental parameter combinations are input one by one into the modified mapping model obtained in S5. The modified mapping model outputs a simulated population growth rate for each set of candidate environmental parameter combinations. When the simulated population growth rate of the candidate environmental parameter combinations generated along the preferred search direction begins to show a downward trend, the search along the preferred search direction is terminated. Taking the candidate environmental parameter combination corresponding to the maximum simulated population growth rate as the center, supplementary candidate environmental parameter combinations are generated in the search space around the center. The supplementary candidate environmental parameter combinations are generated by performing a grid search near the center with a step size smaller than the preset step size. The supplementary candidate environmental parameter combinations are also input into the modified mapping model to obtain the corresponding simulated population growth rate.

[0045] Candidate environmental parameter combinations whose simulated population growth rate meets the preset population growth rate condition are output as propagation conditions. The preset population growth rate condition is used to screen environmental parameter combinations that can meet the propagation target. The preset population growth rate condition is set according to the propagation production target; for example, it can be set as the maximum value of the simulated population growth rate corresponding to all candidate environmental parameter combinations, or as the minimum population growth rate value required by the propagation production plan. All candidate environmental parameter combinations and supplementary candidate environmental parameter combinations are traversed, and candidate environmental parameter combinations whose simulated population growth rate meets the preset population growth rate condition are output as propagation conditions. Propagation conditions are a set of defined temperature, humidity, and substrate ratio values. During the generation of candidate environmental parameter combinations, if the simulated population growth rate of multiple candidate environmental parameter combinations meets the preset population growth rate condition, the candidate environmental parameter combination with the largest simulated population growth rate is selected as the output propagation condition. If the simulated population growth rate of multiple candidate environmental parameter combinations meets the preset population growth rate condition and the simulated population growth rates are equal, the candidate environmental parameter combination with the closest Euclidean distance to the search starting point is selected as the output propagation condition. The output propagation conditions provide numerical references for setting environmental parameters in subsequent propagation stages of the current batch, guiding the propagation facility to adjust the temperature, humidity, and substrate ratio to the corresponding values ​​in the propagation conditions.

[0046] Example 2: Figure 2 A schematic diagram of the data-modeling-based system for the propagation of wood-eating insect populations is provided. The data-modeling-based system for the propagation of wood-eating insect populations includes: The initial modeling module is used to acquire environmental parameters and population response data from multiple propagation batches of wood-eating insects and establish an initial mapping model of population growth rate relative to environmental parameters. The residual determination module is used to obtain the predicted population response and calculate the residual sequence based on the environmental parameters and population response data of the current batch to be established for propagation conditions, using the initial mapping model to determine the residual sequence, dividing the continuous time window and determining whether the residual sequence has a non-random structure. The latent variable representation module is used to extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window when a non-random structure exists, and to use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as the latent variable representation. The effective filtering module is used to extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. The model correction module is used to correct the initial mapping model by utilizing the effective latent variables to obtain the corrected mapping model. The condition optimization module is used to simulate and optimize the population growth rate using the environmental parameters of the current batch to be optimized as the search space and the modified mapping model. It outputs the combination of environmental parameters that meet the population growth rate conditions as the propagation conditions.

[0047] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0048] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0049] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0050] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0051] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0052] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0053] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0054] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0055] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0056] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining the population propagation conditions of wood-eating insects based on data modeling, characterized in that, include: S1: Obtain environmental parameters and population response data for multiple propagation batches of wood-eating insects, and establish an initial mapping model of population growth rate relative to environmental parameters; S2: For the current batch whose propagation conditions need to be established, the predicted population response is obtained and the residual sequence is calculated using the initial mapping model based on the environmental parameters and population response data of its initial stage. The continuous time window is divided and it is determined whether the residual sequence has a non-random structure. S3: When a non-random structure exists, extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window, and use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as a latent variable to represent its effect. S4: Extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. S5: Use the effect of effective latent variables to characterize the modified initial mapping model and obtain the modified mapping model; S6: Using the environmental parameters of the current batch to be optimized as the search space, the modified mapping model is used to simulate and optimize the population growth rate, and the combination of environmental parameters that meets the population growth rate condition is output as the propagation condition.

2. The method for determining the conditions for the propagation of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S1 includes: The environmental parameters and population response data of each batch in multiple propagation batches were obtained at the sampling time points during the propagation cycle. The environmental parameters included temperature, humidity, and substrate ratio, and the population response data included the population size at each sampling time point. For each batch, the instantaneous population growth rate is calculated from the population size at adjacent sampling time points. The instantaneous population growth rate is then aligned with the environmental parameters at the corresponding sampling time points to form a sample set of environmental parameters and population growth rate. Multiple batches of environmental parameters and population growth rate samples are merged, and multiple regression fitting is performed on the merged sample set to obtain a regression function with population growth rate as the response variable and environmental parameters as the explanatory variables. The regression function is used as the initial mapping model.

3. The method for propagation conditions of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S2 include: Input the environmental parameters of the initial stage of the current batch into the initial mapping model, output the predicted population response, and calculate the difference between the actual population response and the predicted population response in the initial stage as the residual sequence. The initial stage is divided into multiple continuous time windows at equal intervals according to the substrate fermentation process. The variance of the residual sequence within each continuous time window is calculated, and the variance is used as the fluctuation amplitude of the residual sequence within that continuous time window. The fluctuation amplitudes of each consecutive time window are arranged in chronological order to obtain the fluctuation amplitude sequence. The monotonicity of the fluctuation amplitude sequence is tested. If the fluctuation amplitude sequence shows a monotonically increasing or monotonically decreasing trend, it is determined that the residual sequence has a non-random structure.

4. The method for determining the conditions for the propagation of wood-eating insect populations based on data modeling according to claim 3, characterized in that, When determining the monotonic trend of the fluctuation amplitude of the residual sequence within each continuous time window as the substrate fermentation process progresses, if the fluctuation amplitude sequence shows a monotonically increasing trend, it is determined that the residual sequence has a non-random structure generated by the strengthening of the cross-interface microbial community as the fermentation process progresses; if the fluctuation amplitude sequence shows a monotonically decreasing trend, it is determined that the residual sequence has a non-random structure generated by the decay of the cross-interface microbial community as the fermentation process progresses.

5. The method for propagation conditions of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S3 include: For each continuous time window, after removing the linear effects of other environmental factors from the residual sequence and each environmental factor sequence, the partial dependence between the residual sequence and a single environmental factor is calculated, thus obtaining the partial dependence of the residual sequence on each environmental factor under each continuous time window. The partial dependencies of the residual sequences corresponding to each environmental factor under each continuous time window are arranged in chronological order to form a point sequence in the feature space spanned by the environmental factors. The movement path of the point sequence in the feature space is taken as the drift trajectory, and the drift trajectory is used as a latent variable to represent the effect.

6. The method for propagation conditions of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S4 include: The trajectory points corresponding to each consecutive time window on the drift trajectory are taken as a sub-representation, resulting in a set of sub-representations with the same number of consecutive time windows; In the feature space spanned by environmental factors, the pairwise distances between each sub-representation in the sub-representation set are calculated, and the mean of all pairwise distances is taken as the concentration of the sub-representation set. The concentration is compared with a preset threshold. If the concentration is higher than the preset threshold, the complete trajectory component corresponding to all sub-representations in the drift trajectory is retained as an effective latent variable action representation.

7. The method for determining the conditions for the propagation of wood-eating insect populations based on data modeling according to claim 6, characterized in that, When comparing the concentration with a preset threshold, the preset threshold is set according to the shape of the drift trajectory of the partial dependence relationship in the feature space under each continuous time window. The more the shape of the drift trajectory is oriented as a straight line, the higher the preset threshold is; the more the shape of the drift trajectory is randomly distributed, the lower the preset threshold is.

8. The method for propagation conditions of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S5 include: The effective latent variable effect representation is expanded into a correction sequence according to the order of each continuous time window, and the effective latent variable effect representation component corresponding to each continuous time window in the correction sequence is used as the output correction amount of the initial mapping model within the corresponding continuous time window. The corrected output is obtained by subtracting the output correction amount of the corresponding continuous time window from the output of the initial mapping model in each continuous time window. The corrected outputs of all continuous time windows are then integrated to obtain the corrected mapping model.

9. The method for propagation conditions of wood-eating insect populations based on data modeling according to claim 1, characterized in that, S6 include: The search space is defined as the allowable adjustment range of environmental parameters in the current batch to be optimized. Extract the drift direction of the partial dependency relationship with continuous time window from the effective latent variable action representation, and map the drift direction to the search space to obtain the priority search direction; Multiple sets of candidate environmental parameter combinations are generated in the search space along the preferred search direction. Each set of candidate environmental parameter combinations is input into the modified mapping model to obtain the simulated population growth rate corresponding to each set of candidate environmental parameter combinations. The candidate environmental parameter combinations that simulate a population growth rate that meets the preset population growth rate conditions are output as propagation conditions.

10. A data-modeling-based system for the propagation of wood-eating insect populations, used to implement the data-modeling-based method for the propagation of wood-eating insect populations as described in any one of claims 1-9, characterized in that, include: The initial modeling module is used to acquire environmental parameters and population response data from multiple propagation batches of wood-eating insects and establish an initial mapping model of population growth rate relative to environmental parameters. The residual determination module is used to obtain the predicted population response and calculate the residual sequence based on the environmental parameters and population response data of the current batch to be established for propagation conditions, using the initial mapping model to determine the residual sequence, dividing the continuous time window and determining whether the residual sequence has a non-random structure. The latent variable representation module is used to extract the partial dependency relationship between the residual sequence and each environmental factor according to each continuous time window when a non-random structure exists, and to use the drift trajectory of the partial dependency relationship under each continuous time window in the feature space as the latent variable representation. The effective filtering module is used to extract sub-representations from the drift trajectory within each continuous time window, and retain the drift trajectory components corresponding to the sub-representations with a concentration higher than a preset threshold as effective latent variable action representations. The model correction module is used to correct the initial mapping model by utilizing the effective latent variables to obtain the corrected mapping model. The condition optimization module is used to simulate and optimize the population growth rate using the environmental parameters of the current batch to be optimized as the search space and the modified mapping model. It outputs the combination of environmental parameters that meet the population growth rate conditions as the propagation conditions.

Images