Construction method of integrated model for bacterial detection

Abstract

The invention relates to the technical field of data model construction, in particular to a construction method of an integrated model for bacterial detection. The method comprises the following steps: acquiring a bacteria original data set; performing data denoising on the original bacteria data set to obtain to-be-detected bacteria data; pre-classifying the to-be-detected bacterial data to obtain pre-classified bacterial data; performing data batch identification on the pre-classified bacterial data to obtain to-be-detected batch data; performing batch deviation detection on the to-be-detected batch data to obtain bacteria batch deviation data; performing batch correction on the to-be-detected bacteria data according to the bacteria batch deviation data to obtain bacteria correction data; through the data processing technology, the feature mapping technology, the intelligent feedback technology and the self-adaptive adjustment technology, the generalization ability of the integrated model for bacteria detection is enhanced, so that the to-be-detected data of bacteria are rapidly detected and responded, and the bacteria detection efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data model construction, and in particular to a method for constructing an integrated model for bacteria detection. Background Art

[0002] Early bacterial detection relied on traditional culture and biochemical identification methods. Although these methods are mature, they are time-consuming. With the development of molecular biology technology, bacterial identification technology based on DNA sequence analysis has gradually emerged, which can provide more accurate bacterial classification information. The emergence of microfluidics has brought revolutionary changes to bacterial detection. Microfluidics technology realizes dynamic monitoring of bacteria and their interaction processes by integrating microfluidic control and a variety of optoelectronic sensing monitoring modules. In recent years, with the development of artificial intelligence technology, machine learning and deep learning algorithms have been widely used in bacterial detection, especially in the processing and analysis of bacterial detection data, specifically through feature selection and model training, and the construction of an integrated model for bacterial detection; identify bacteria from high-dimensional and complex bacterial community data. However, the existing integrated model for bacterial detection has low generalization ability for complex bacterial detection data, and it is difficult to achieve rapid detection response, resulting in low efficiency and accuracy of bacterial detection. Summary of the invention

[0003] Based on this, it is necessary to provide a method for constructing an integrated model for bacterial detection to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for constructing an integrated model for bacteria detection is provided, the method comprising the following steps:

[0005] Step S1: obtaining a bacteria original data set; performing data denoising on the bacteria original data set to obtain bacteria data to be detected; performing pre-classification on the bacteria data to be detected to obtain pre-classified bacteria data; performing data batch identification on the pre-classified bacteria data to obtain batch data to be detected; performing batch deviation detection on the batch data to be detected to obtain bacteria batch deviation data; performing batch correction on the bacteria data to be detected according to the bacteria batch deviation data to obtain bacteria correction data;

[0006] Step S2: performing bacterial characteristic recognition on the pre-classified bacterial data according to the bacterial correction data to obtain bacterial characteristic data; performing bacterial behavior pattern determination on the bacterial characteristic data to obtain bacterial behavior pattern data; performing bacterial type characteristic recognition on the pre-classified bacterial data according to the bacterial behavior pattern data to generate type characteristic data; performing type characteristic mapping on the bacterial data to be detected based on the type characteristic data to obtain bacterial type mapping data;

[0007] Step S3: constructing a bacteria detection integrated model for bacteria data to be detected based on bacteria correction data and bacteria type mapping data to obtain a bacteria detection integrated pre-model; enhancing the generalization capability of the bacteria detection integrated pre-model to obtain model generalization enhancement measures; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measures to obtain adaptive generalization adjustment data; performing rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model;

[0008] Step S4: Perform performance evaluation on the bacteria detection integrated training model to generate training model performance evaluation data; perform intelligent iterative optimization on the bacteria detection integrated training model based on the training model performance evaluation data to obtain a bacteria detection integrated model; perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

[0009] The present invention significantly improves the quality and accuracy of the original bacterial data set through data denoising, ensuring the effectiveness of subsequent analysis; pre-classification processing allows bacterial data to be grouped according to specific features; data batch identification ensures that data from different batches can be considered independently, thereby reducing the impact of batch differences on analysis results; batch deviation detection further reveals the specific differences between batches, providing a basis for batch correction; batch correction guided by bacterial batch deviation data effectively reduces batch variation and improves data consistency and reliability. The use of bacterial correction data makes bacterial characteristic identification more accurate and provides accurate basic data for subsequent bacterial behavior pattern analysis; the determination of bacterial behavior patterns clarifies the biological characteristics and potential behavioral laws of bacteria; the generation of type feature data provides clear standards for bacterial classification, making bacterial type feature identification more accurate. The combination of bacteria correction data and bacteria type mapping data provides a data basis for building a bacteria detection integrated model, ensuring the accuracy and reliability of the model; the generalization ability of the integrated model is enhanced, and the detection ability of the model for unknown data is improved, thereby enhancing the practicality of the model; the adaptive generalization adjustment further optimizes the generalization performance of the model, ensuring that the model can maintain stable performance on different data sets; the bacteria detection integrated pre-model is trained for rapid detection response integration, so that the model can quickly adapt to new data, improving the response speed and detection efficiency of the model. Performance evaluation provides an objective evaluation standard for the bacteria detection integrated training model, ensuring the optimization and improvement points of the model; intelligent iterative optimization is based on performance evaluation data, further improving the performance of the model, ensuring the efficiency and accuracy of the model in practical applications; finally, the application of the bacteria detection integrated model realizes the accurate detection of bacteria to be detected data, generates a detailed bacteria detection report, and provides strong support for the identification, classification and research of bacteria. Therefore, the present invention enhances the generalization ability of the integrated model of bacteria detection through data processing technology, feature mapping technology, intelligent feedback technology and adaptive adjustment technology, so that the bacteria to be detected data can be quickly detected and responded, thereby improving the efficiency and accuracy of bacteria detection.

[0010] Preferably, step S2 comprises the following steps:

[0011] Step S21: applying multi-parameter statistics to the bacterial correction data to obtain bacterial statistical data; extracting bacterial biomarkers from the bacterial statistical data to obtain bacterial biomarker data; and identifying bacterial characteristics of the pre-classified bacterial data based on the bacterial biomarker data to obtain bacterial characteristic data;

[0012] Step S22: grouping the bacterial characteristic data to obtain bacterial characteristic grouping data; identifying bacterial groups according to the bacterial characteristic grouping data to obtain bacterial group data; dividing the bacterial group data into performance behaviors to generate bacterial behavior division data; determining the behavior pattern of the bacterial characteristic data according to the bacterial behavior division data to obtain bacterial behavior pattern data;

[0013] Step S23: extracting multi-mode standard rules from the bacterial behavior pattern data to generate bacterial multi-mode standard rules; comparing the pattern rules of the pre-classified bacterial data one by one according to the bacterial multi-mode standard rules to obtain bacterial key feature data; identifying bacterial type features from the bacterial key feature data to generate type feature data;

[0014] Step S24: Associating bacteria data to be detected with bacteria types through type feature data to obtain bacteria type association data; quantifying the degree of association of the bacteria type association data to generate type association quantification data; assigning type labels to the bacteria data to be detected based on the type association quantification data to obtain labels of the data to be detected; performing type feature mapping on the labels of the data to be detected based on the type feature data to obtain bacteria type mapping data.

[0015] The present invention can reveal the statistical laws and distribution characteristics in the data by applying multi-parameter statistics to bacterial correction data; extract bacterial biomarkers from bacterial statistical data to clearly identify bacterial data with biologically significant indicators; use bacterial biomarker data to perform characteristic recognition on pre-classified bacterial data to accurately identify the characteristics of bacteria; group bacterial characteristic data to classify bacteria with similar characteristics; identify bacterial populations based on bacterial characteristic grouping data to divide bacteria into different bacterial populations; further, perform performance behavior classification on bacterial population data to understand the behavioral characteristics of different bacterial populations; determine the behavioral patterns of bacterial characteristic data based on bacterial behavioral classification data , which can reveal the behavioral laws of bacteria; the extraction of bacterial behavior pattern data can be used to obtain multi-modal standard rules; the pre-classified bacterial data are compared with the pattern rules one by one according to the bacterial multi-modal standard rules to clearly identify the key characteristics of bacteria; the bacterial data to be detected are associated with bacterial types through type feature data, so that the bacterial types of the bacterial data to be detected are associated with feature types; the degree of association of bacterial type association data is quantified, providing a quantitative evaluation of bacterial type association; type labels are assigned to bacterial data to be detected according to type association quantification data, so as to realize rapid identification of bacterial types; type feature mapping is performed on the labels of the data to be detected based on the type feature data, which provides an intuitive mapping for the classification of bacteria.

[0016] Preferably, step S3 comprises the following steps:

[0017] Step S31: performing uniform normalization on the bacteria correction data to obtain uniform bacteria correction data; performing mapping enhancement on the bacteria type mapping data to generate type mapping enhancement data; demarcating the bacteria to-be-detected sample groups based on the bacteria uniform correction data and the type mapping enhancement data to generate to-be-detected sample group data; performing type feature correspondence on the to-be-detected sample group data according to the type mapping enhancement data to generate type feature correspondence data; performing type feature label extraction on the type feature correspondence data to generate type feature label data;

[0018] Step S32: performing bacterial colony feature recognition on the bacterial uniformity correction data to obtain bacterial colony feature data; performing colony distribution detection on the bacterial colony feature data to obtain bacterial colony distribution data; performing colony type number counting on the type feature corresponding data through the bacterial colony distribution data to obtain colony type number data; performing colony number label extraction on the colony type number data to generate colony number label data;

[0019] Step S33: performing bacterial growth state analysis on the bacterial colony characteristic data to obtain bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data; constructing a bacterial detection integrated model for the bacteria to be detected data according to the type characteristic label data, the colony quantity label data and the colony state label data to obtain a bacterial detection integrated pre-model;

[0020] Step S34: enhancing the generalization capability of the bacteria detection integrated pre-model to obtain a model generalization enhancement measure; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain adaptive generalization adjustment data;

[0021] Step S35: Perform rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model.

[0022] The present invention enhances the consistency and comparability of bacterial correction data through data uniform normalization processing, providing a standardized basis for subsequent analysis; mapping enhancement processing improves the discrimination of bacterial type mapping data, making the identification of bacterial types more accurate; combining bacterial uniform correction data and type mapping enhancement data, the bacterial samples to be detected are grouped to ensure the accuracy of sample classification; type feature correspondence is performed on the group data of the samples to be detected according to the type mapping enhancement data, so that the type mapping enhancement data establishes a direct connection with the group data of the samples to be detected, providing a basis for feature extraction; type feature label extraction further clarifies the key attributes of bacterial samples; bacterial colony feature recognition improves the ability to recognize bacterial colony morphology, and colony distribution detection clarifies the spatial distribution of bacterial colonies, providing a basis for counting the number of colony types; Species counting makes it possible to quantify the types and quantities of different colonies, and colony quantity label extraction clarifies the quantitative characteristics of colonies; bacterial growth state analysis helps to evaluate bacterial growth, and colony state label extraction provides key information on colony growth state for subsequent model construction, thereby enhancing the model's ability to identify bacterial growth; type feature label data, colony quantity label data, and colony state label data are integrated to form a bacterial detection framework; the generalization ability of the bacterial detection integrated pre-model is enhanced, which improves the model's adaptability to unknown data and detection accuracy; adaptive generalization adjustment further optimizes the generalization performance of the model and ensures the stability and reliability of the model on different data sets; rapid detection response integrated training enables the bacterial detection integrated training model to efficiently respond to new detection needs, improving the application efficiency and practicality of the model.

[0023] Preferably, step S33 includes the following steps:

[0024] Step S331: performing colony contour recognition on the bacterial colony characteristic data to generate bacterial colony contour data; performing morphological measurement on the bacterial colony contour data to obtain bacterial morphological data;

[0025] Step S332: performing a colony two-dimensional morphological image scanning on the bacterial colony characteristic data based on the bacterial morphological data to generate a colony two-dimensional morphological image; performing a colony three-dimensional morphological reconstruction on the colony two-dimensional morphological image to obtain three-dimensional colony morphological data;

[0026] Step S333: performing bacterial volume detection on the three-dimensional colony morphology data to obtain bacterial volume data; performing bacterial division identification on the three-dimensional colony morphology data to obtain bacterial division state data; determining bacterial growth state of the bacteria to be detected data according to the bacterial division state data and the bacterial volume data to generate bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data;

[0027] Step S334: using the type feature label data, colony quantity label data and colony status label data as a bacteria detection integrated model classifier component to obtain a model classifier component; inputting the to-be-detected sample group data into the model classifier component to obtain detection sample classification data;

[0028] Step S335: Perform detection integration and fusion on the group data of the samples to be detected according to the detection sample classification data to obtain a bacteria detection integrated pre-model.

[0029] The present invention accurately depicts the appearance characteristics of bacterial colonies through colony contour recognition; performs morphological measurement on bacterial colony contour data to further refine the morphological characteristics of bacterial colonies for subsequent identification and differentiation of different bacterial types; performs two-dimensional morphological image scanning using bacterial morphological data to intuitively display the planar characteristics of colonies and provide necessary image information for further three-dimensional morphological reconstruction; performs three-dimensional morphological reconstruction of colonies on two-dimensional morphological images of colonies to enable a more comprehensive understanding of the growth status of colonies; performs volume detection on three-dimensional colony morphological data to quantify the volume of bacterial colonies and provide volume parameters for evaluating the growth status of colonies; clarifies the reproduction status of bacteria by identifying the bacterial division status; combines bacterial division with the The state and volume data can accurately determine the growth state of bacteria, and then generate bacterial growth state data; extract colony state label data from bacterial growth state data to provide a clear classification basis for subsequent bacterial detection; integrate type feature label data, colony number label data and colony state label data, and use different types of data to improve the classification accuracy of the model; input the group data of the samples to be tested into the model classifier component to achieve accurate classification of the samples; integrate and fuse the group data of the samples to be tested according to the classification data of the tested samples to construct an integrated pre-model for bacterial detection; by integrating multiple test results, the reliability and stability of the test are improved, and a pre-model is provided for the final bacterial detection.

[0030] Preferably, step S34 includes the following steps:

[0031] Step S341: determining the classification feature layer of the bacteria detection integrated pre-model to obtain classification feature layer information; performing feature structure recognition on the classification feature layer information to obtain classification feature structure data; performing feature structure slicing on the model classifier component according to the classification feature structure data to generate feature structure slicing data; performing type range segmentation on the type feature label data through the feature structure slicing data to obtain type segmentation data;

[0032] Step S342: performing colony magnitude gradient grouping on the colony quantity label data according to the type segmentation data to obtain colony magnitude gradient data; performing colony state cycle determination on the colony state label data according to the colony magnitude gradient data to obtain colony state cycle data;

[0033] Step S343: performing multi-dimensional feature fusion on the type subdivision data, colony magnitude gradient data and colony state cycle data to generate multi-dimensional feature fusion data; performing feature autoencoding compression on the multi-dimensional feature fusion data to obtain multi-dimensional feature compression data;

[0034] Step S344: performing feature multi-scale recognition on the multi-dimensional feature compression data to generate multi-scale feature data; performing detection label weight extraction on the model classifier component to obtain detection weight data; performing feature weight mapping on the detection weight data through the multi-scale feature data to generate multi-scale detection weight data; performing dynamic weight generalization enhancement on the bacteria detection integrated pre-model through the multi-scale detection weight data to obtain model generalization enhancement measures;

[0035] Step S345: adjusting the detection priority of the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain priority adjustment data; and performing adaptive learning rate optimization on the priority adjustment data to obtain adaptive generalization adjustment data.

[0036] The present invention determines the classification feature layer of the integrated pre-model for bacteria detection and clarifies the key information used for classification in the model, thereby improving the accuracy of classification; identifies the structure of the classification feature layer information and can understand the relationship between features; performs feature structure slicing on the model classifier component based on the classification feature structure data and can refine the type feature label data, thereby achieving accurate segmentation of the type range; uses the type segmentation data to perform gradient grouping on the colony quantity label data and can effectively classify the colony quantity according to magnitude; determines the colony state cycle according to the colony magnitude gradient data, which helps to grasp the growth and change of the colony; multi-dimensionally integrates the type segmentation data, the colony magnitude gradient data and the colony state cycle data It can integrate feature information from different sources to form comprehensive multi-dimensional feature data; reduce data dimensions and improve data processing efficiency through feature autoencoding compression; perform multi-scale recognition on multi-dimensional feature compressed data to extract feature information at different levels and improve the model's recognition ability for complex data; extract detection label weights to help identify key features in the model; optimize the model's weight distribution through feature weight mapping, thereby enhancing the model's generalization ability; adjust detection priorities based on model generalization enhancement measures, optimize the model's detection process, and improve detection efficiency; further improve the model's generalization performance through adaptive learning rate optimization, so that the model can better adapt to new bacteria to be detected data.

[0037] Preferably, step S345 includes the following steps:

[0038] Step S3451: performing non-linear feature screening on the multi-scale detection weight data according to the model generalization enhancement measure to generate non-linear feature screening data; performing detection contribution calculation on the detection sample classification data according to the non-linear feature screening data to obtain detection contribution data;

[0039] Step S3452: performing contribution level division on the detection contribution data to obtain contribution level level data; performing detection weight sorting on the detection weight data based on the contribution level level data to generate weight sorting data; performing detection priority weighted calculation on the weight sorting data to obtain detection priority weighted data;

[0040] Step S3453: assigning multi-scale priority to the detection weight data according to the detection priority weighted data and the multi-scale feature data to obtain multi-scale priority assignment information; performing detection priority transformation on the multi-scale detection weight data based on the multi-scale priority assignment information to obtain priority adjustment data;

[0041] Step S3454: performing dynamic feature sensitivity detection on the priority adjustment data to obtain feature sensitivity data; performing sensitivity level classification on the weight sorting data according to the feature sensitivity data to obtain weight sensitivity level data; performing weight feature learning rate calculation on the weight sensitivity level data to generate feature learning rate data;

[0042] Step S3455: dynamically divide the feature learning rate data according to the detection priority weighted data to obtain learning rate division data; and adaptively optimize the priority adjustment data according to the dynamic learning rate division to obtain adaptive generalization adjustment data.

[0043] The present invention identifies nonlinear features that have a significant impact on the classification of detection samples through nonlinear feature screening, thereby improving the classification accuracy of the model; detection contribution measurement can quantify the contribution of each feature to the classification result; contribution level division helps to distinguish the contribution of different features to the detection result, thereby sorting the detection weights; through weighted calculation, reasonable priorities are assigned to detection samples to ensure that key samples are given priority; multi-scale priority assignment can reasonably assign detection weights according to the importance of features, and priority transformation ensures that the detection process can be dynamically adjusted according to the importance of features, thereby optimizing detection efficiency; dynamic feature sensitivity detection enables identification of features that are most sensitive to detection results, and sensitivity level division further refines the sensitivity of features; weighted feature learning rate calculation assigns appropriate learning rates to features of different sensitivities to optimize the learning process of the model; dynamic learning rate division ensures that feature learning rate data can be reasonably assigned according to detection priority weighted data, and adaptive learning rate optimization further improves the generalization ability and learning efficiency of the model.

[0044] Preferably, step S35 includes the following steps:

[0045] Step S351: performing single feature matching on the model classifier component according to the adaptive generalization adjustment data to generate single feature matching data; performing multi-scale feature association matching on the sample group data to be detected based on the single feature matching data to obtain multi-scale feature association data;

[0046] Step S352: performing feature detection interconnection on the multi-dimensional feature fusion data according to the multi-scale feature association data to obtain detection feature interconnection data; performing label parameter extraction on the detection feature interconnection data to generate interconnection label parameters;

[0047] Step S353: performing detection priority interconnection on the detection priority weighted data based on the interconnection tag parameters to obtain detection priority interconnection data; performing detection contribution increment on the detection priority interconnection data to obtain detection contribution increment data; performing detection weight re-ranking on the weight ranking data according to the detection contribution increment data to generate weight ranking response data;

[0048] Step S354: Perform a fast learning rate response on the learning rate division data according to the weight sorting response data to obtain the learning rate fast response data; perform a fast pre-detection on the sample group data to be detected through the learning rate fast response data to obtain the sample pre-detection data; input the sample pre-detection data into the bacteria detection integrated pre-model for fast detection response training to obtain the bacteria detection integrated training model.

[0049] The present invention uses adaptive generalization to adjust data for single feature matching of model classifier components, thereby ensuring the precise correspondence between features and models and improving the accuracy of matching; multi-scale feature association matching further enhances the correlation between the sample group data to be detected and the model; feature detection interconnection realizes the deep integration of multi-dimensional feature fusion data through multi-scale feature association data and enhances the mutual connection between features; label parameter extraction provides key parameter information for feature detection and helps to optimize the detection process; detection priority interconnection ensures that the detection process can be dynamically adjusted according to the interconnected label parameters and improves the priority management of detection; the detection contribution increment reflects the new contribution of the feature to the detection result, and the generation of weighted ranking response data helps to adjust the detection weight to adapt to new detection requirements; the generation of learning rate fast response data speeds up the adjustment speed of the learning rate and improves the adaptability of the model to new data; rapid pre-detection provides preliminary detection results for the sample group data to be detected, and after inputting the bacterial detection integrated pre-model, a bacterial detection integrated training model is finally formed through rapid detection response training, thereby improving the detection efficiency and accuracy of the model.

[0050] Preferably, step S4 comprises the following steps:

[0051] Step S41: Divide the to-be-detected sample group data into a test set and a validation set to obtain a to-be-detected sample test set and a to-be-detected sample validation set;

[0052] Step S42: input the test set of samples to be detected into the bacteria detection integrated training model to measure the detection accuracy and generate test set accuracy data; input the test set of samples to be detected into the bacteria detection integrated training model to record the recall rate and generate test set recall rate data; integrate the test set accuracy rate data and the test set recall rate data to obtain the test set performance data;

[0053] Step S43: input the verification set of samples to be tested into the bacteria detection integrated training model to measure the detection accuracy and generate verification set accuracy data; input the verification set of samples to be tested into the bacteria detection integrated training model to record the recall rate and generate verification set recall rate data; integrate the verification set accuracy rate data and the verification set recall rate data to obtain the verification set performance data;

[0054] Step S44: performing a performance comparison between the test set performance data and the validation set performance data to generate training model performance evaluation data; iteratively optimizing the bacteria detection integrated training model according to the training model performance evaluation data to obtain a bacteria detection integrated model;

[0055] Step S45: Perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

[0056] The present invention ensures the comprehensiveness and accuracy of model evaluation by dividing the sample group data to be detected into a test set and a validation set, and respectively using them for performance evaluation and validation of the model; inputting the test set into the bacteria detection integrated training model, calculating the accuracy and recall rate, and generating test set performance data; this helps to evaluate the performance of the model in practical applications, and ensure the accuracy and comprehensiveness of the model's detection results; inputting the validation set into the bacteria detection integrated training model, and similarly calculating the accuracy and recall rate, and generating validation set performance data, which helps to further confirm the stability and reliability of the model; by comparing the performance data of the test set and the validation set, the performance of the model can be comprehensively evaluated, and a basis can be provided for iterative optimization of the model; the model is optimized according to the performance evaluation data, thereby improving the accuracy and stability of the model; the optimized bacteria detection integrated model is used to detect the bacteria data to be detected, and a bacteria detection report is generated, thereby ensuring the accuracy and reliability of the detection results.

[0057] Preferably, step S44 includes the following steps:

[0058] Step S441: perform similarity measurement on the test set performance data and the validation set performance data to generate performance similarity data; perform similarity deviation calculation on the performance similarity data to obtain performance similarity deviation data; perform detection response time difference calculation on the test set performance data and the validation set performance data to obtain response time difference data;

[0059] Step S442: performing performance evaluation on the integrated training model for bacteria detection through performance similarity deviation data and response time difference data, and generating training model performance evaluation data;

[0060] Step S443: performing anomaly detection feedback on the bacteria detection integrated training model according to the training model performance evaluation data to obtain anomaly detection feedback data; performing an association parameter improvement operation on the multi-scale feature association data according to the anomaly detection feedback data to generate association parameter improvement data;

[0061] Step S444: performing error back propagation on the performance similarity deviation data according to the associated parameter improvement data to obtain error back propagation data; performing loss function back propagation on the detection priority weighted data according to the error back propagation data to obtain loss function back propagation data; performing regularization constraints on the bacteria detection integrated training model according to the loss function back propagation data to obtain regularization constraint data;

[0062] Step S445: performing multi-task collaborative training on the sample pre-detection data according to the regularized constraint data to obtain multi-task collaborative data; iteratively optimizing the bacteria detection integrated training model according to the multi-task collaborative data to obtain the bacteria detection integrated model.

[0063] The present invention measures the similarity between the test set performance data and the validation set performance data, and calculates the similarity deviation to evaluate the performance consistency of the model on different data sets; at the same time, the detection response time difference is calculated to clarify the response of the model on different data sets; the performance similarity deviation data and the response time difference data are combined to evaluate the performance of the bacteria detection integrated training model, comprehensively evaluate the accuracy and efficiency of the model, generate comprehensive performance evaluation data, and provide a basis for model optimization; anomaly detection feedback is performed based on the performance evaluation data to identify the abnormal performance of the model under specific circumstances; and the multi-scale feature association data is associated with the parameter improvement operation, so as to improve the model's ability to identify abnormal situations; the model parameters are adjusted through error back propagation to reduce performance deviation; the loss function back propagation is helpful to optimize the detection weight and improve the prediction accuracy of the model; the application of regularization constraints prevents the model from overfitting and ensures that the model has good generalization ability; multi-task collaborative training improves the performance of the model on different tasks and enhances the generalization ability of the model; the bacteria detection integrated training model is iteratively optimized to further improve the performance of the model and make it more stable and reliable.

[0064] Preferably, step S444 includes the following steps:

[0065] Step S4441: determining the parameter improvement range of the associated parameter improvement data to obtain associated parameter range data; calculating the deviation degree of the performance similarity deviation data to obtain performance deviation degree data;

[0066] Step S4442: mapping the associated parameter range data with the performance deviation degree data to generate deviation parameter mapping data; performing error impact identification on the deviation parameter mapping data to obtain error impact data;

[0067] Step S4443: performing error back propagation on the multi-scale feature association data according to the error influence data to obtain error back propagation data; performing detection network layer extraction on the bacteria detection integrated training model to obtain detection network layer data; performing loss function calculation layer by layer on the detection network layer data according to the detection priority weighted data to obtain the partial derivative of the loss function;

[0068] Step S4444: The partial derivative of the loss function is transmitted back to the detection network layer to obtain the loss function back-propagation data; the regularization function is determined according to the anomaly detection feedback data to obtain the regularization function; the regularization function is added to the loss function back-propagation data to generate the back-propagation function information; the bacteria detection integrated training model is regularized and constrained according to the back-propagation function information to obtain the regularization constraint data.

[0069] The present invention can accurately control the amplitude of parameter adjustment and ensure the stability of model optimization by determining the range of associated parameter improvement; the measurement of performance deviation can quantify the deviation degree of model performance; the associated parameter range is mapped with the performance deviation degree, and the key parameters in the model are accurately identified and adjusted; the error impact identification further refines the specific influencing factors of the model performance deviation; the error back propagation uses the error impact data to adjust the multi-scale feature association data, which can reduce the error of model detection; the extraction of the detection network layer provides structured data for the in-depth analysis of the model; the loss function is calculated layer by layer to ensure that each layer of the model can be properly optimized; the reverse feedback of the partial derivative of the loss function enables the network layer of the model to be effectively adjusted according to the error; the regularization function is determined and added to the loss function, so as to enhance the resistance of the model to overfitting; the application of the regularization constraint further improves the generalization ability of the model and ensures that the model can maintain stable performance on different data sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A schematic diagram of the steps of a method for constructing an integrated model for bacterial detection;

[0071] Figure 2for Figure 1 Detailed implementation steps of step S3 in FIG.

[0072] Figure 3 for Figure 2 Detailed implementation steps of step S33;

[0073] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0074] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0075] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0076] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0077] To achieve this, please refer to Figures 1 to 3 , a method for constructing an integrated model for bacteria detection, the method comprising the following steps:

[0078] Step S1: obtaining a bacteria original data set; performing data denoising on the bacteria original data set to obtain bacteria data to be detected; performing pre-classification on the bacteria data to be detected to obtain pre-classified bacteria data; performing data batch identification on the pre-classified bacteria data to obtain batch data to be detected; performing batch deviation detection on the batch data to be detected to obtain bacteria batch deviation data; performing batch correction on the bacteria data to be detected according to the bacteria batch deviation data to obtain bacteria correction data;

[0079] Step S2: performing bacterial characteristic recognition on the pre-classified bacterial data according to the bacterial correction data to obtain bacterial characteristic data; performing bacterial behavior pattern determination on the bacterial characteristic data to obtain bacterial behavior pattern data; performing bacterial type characteristic recognition on the pre-classified bacterial data according to the bacterial behavior pattern data to generate type characteristic data; performing type characteristic mapping on the bacterial data to be detected based on the type characteristic data to obtain bacterial type mapping data;

[0080] Step S3: constructing a bacteria detection integrated model for bacteria data to be detected based on bacteria correction data and bacteria type mapping data to obtain a bacteria detection integrated pre-model; enhancing the generalization capability of the bacteria detection integrated pre-model to obtain model generalization enhancement measures; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measures to obtain adaptive generalization adjustment data; performing rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model;

[0081] Step S4: Perform performance evaluation on the bacteria detection integrated training model to generate training model performance evaluation data; perform intelligent iterative optimization on the bacteria detection integrated training model based on the training model performance evaluation data to obtain a bacteria detection integrated model; perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

[0082] The present invention significantly improves the quality and accuracy of the original bacterial data set through data denoising, ensuring the effectiveness of subsequent analysis; pre-classification processing allows bacterial data to be grouped according to specific features; data batch identification ensures that data from different batches can be considered independently, thereby reducing the impact of batch differences on analysis results; batch deviation detection further reveals the specific differences between batches, providing a basis for batch correction; batch correction guided by bacterial batch deviation data effectively reduces batch variation and improves data consistency and reliability. The use of bacterial correction data makes bacterial characteristic identification more accurate and provides accurate basic data for subsequent bacterial behavior pattern analysis; the determination of bacterial behavior patterns clarifies the biological characteristics and potential behavioral laws of bacteria; the generation of type feature data provides clear standards for bacterial classification, making bacterial type feature identification more accurate. The combination of bacteria correction data and bacteria type mapping data provides a data basis for building a bacteria detection integrated model, ensuring the accuracy and reliability of the model; the generalization ability of the integrated model is enhanced, and the detection ability of the model for unknown data is improved, thereby enhancing the practicality of the model; the adaptive generalization adjustment further optimizes the generalization performance of the model, ensuring that the model can maintain stable performance on different data sets; the bacteria detection integrated pre-model is trained for rapid detection response integration, so that the model can quickly adapt to new data, improving the response speed and detection efficiency of the model. Performance evaluation provides an objective evaluation standard for the bacteria detection integrated training model, ensuring the optimization and improvement points of the model; intelligent iterative optimization is based on performance evaluation data, further improving the performance of the model, ensuring the efficiency and accuracy of the model in practical applications; finally, the application of the bacteria detection integrated model realizes the accurate detection of bacteria to be detected data, generates a detailed bacteria detection report, and provides strong support for the identification, classification and research of bacteria. Therefore, the present invention enhances the generalization ability of the integrated model of bacteria detection through data processing technology, feature mapping technology, intelligent feedback technology and adaptive adjustment technology, so that the bacteria to be detected data can be quickly detected and responded, thereby improving the efficiency and accuracy of bacteria detection.

[0083] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a method for constructing an integrated model for bacteria detection of the present invention. In this example, the method for constructing an integrated model for bacteria detection includes the following steps:

[0084] Step S1: obtaining a bacteria original data set; performing data denoising on the bacteria original data set to obtain bacteria data to be detected; performing pre-classification on the bacteria data to be detected to obtain pre-classified bacteria data; performing data batch identification on the pre-classified bacteria data to obtain batch data to be detected; performing batch deviation detection on the batch data to be detected to obtain bacteria batch deviation data; performing batch correction on the bacteria data to be detected according to the bacteria batch deviation data to obtain bacteria correction data;

[0085] In an embodiment of the present invention, a bacterial sample is obtained by laboratory culture or field sampling, and a high-throughput sequencing technology, such as an Illumina platform, is used to extract and sequence DNA from the sample to obtain a bacterial original data set; the bacterial original data set is quality controlled, and low-quality sequences are detected and removed using tools such as FastQC; the sequence is trimmed using software such as Trimmomatic to remove linker sequences, low-quality tails, and sequences containing N to obtain bacterial data to be detected; the processed data is operated using QIIME or Mothur software, specifically including the selection of OTUs and preliminary classification of bacterial species to obtain pre-classified bacterial data; a statistical method, such as ANOVA or t-SNE, is used to specifically count the batch effect of the batch data to be detected, and the batch deviation is detected by comparing the statistical indicators of samples between different batches, such as α diversity and β diversity; a batch correction algorithm, such as SWARM or Harmony, is used to correct the data to eliminate the batch effect; the corrected data is statistically analyzed again to verify the removal of the batch effect; the bacterial data to be detected is batch corrected according to the bacterial batch deviation data to obtain bacterial corrected data.

[0086] Step S2: performing bacterial characteristic recognition on the pre-classified bacterial data according to the bacterial correction data to obtain bacterial characteristic data; performing bacterial behavior pattern determination on the bacterial characteristic data to obtain bacterial behavior pattern data; performing bacterial type characteristic recognition on the pre-classified bacterial data according to the bacterial behavior pattern data to generate type characteristic data; performing type characteristic mapping on the bacterial data to be detected based on the type characteristic data to obtain bacterial type mapping data;

[0087] In an embodiment of the present invention, bacterial species annotation is performed by bioinformatics analysis software, such as MEGAN or Krona; machine learning algorithms, such as random forests or support vector machines, are applied to analyze the gene sequences of bacteria to identify their biological characteristics and biological markers; bacterial characteristic data are dynamically monitored by time series analysis, such as principal component analysis (PCA), to determine their bacterial behavior patterns; bacterial simulation technology, such as cellular automaton simulation technology, is used to simulate the growth and diffusion behavior of bacteria under different environmental conditions; multi-locus sequence typing (MLST) or whole genome sequencing (WGS) technology is used to identify bacterial type characteristics of pre-classified bacterial data, and identify the molecular typing and evolutionary relationship of bacteria; specific genes between different bacterial types are identified by comparative genomics analysis; type characteristic data are homology analyzed by bioinformatics tools, such as BLAST or Mauve; and the data to be detected are compared and mapped with known bacterial types to obtain bacterial type mapping data.

[0088] Step S3: constructing a bacteria detection integrated model for bacteria data to be detected based on bacteria correction data and bacteria type mapping data to obtain a bacteria detection integrated pre-model; enhancing the generalization capability of the bacteria detection integrated pre-model to obtain model generalization enhancement measures; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measures to obtain adaptive generalization adjustment data; performing rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model;

[0089] In the embodiment of the present invention, a machine learning algorithm, such as a random forest or a gradient boosting machine (GBM), is used to combine bacterial correction data and bacterial type mapping data to build an integrated pre-model for bacterial detection; a feature selection technique, such as principal component analysis (PCA) or an autoencoder, is used to extract the features that are most helpful for bacterial detection from the data, and to reduce the model complexity and improve the detection performance; data enhancement techniques, such as rotation, scaling, color transformation, etc., are used to transform the training data, and specifically, the training data set is rotated (20 degrees), scaled (ratio of 1.2), and color transformed (saturation increased by 20%) and other enhancement operations are performed; a meta-learning strategy is adopted to train on multiple different bacterial data sets, specifically on 5 different bacterial Meta-training is performed on the dataset, and K-fold cross validation (K=5) is used for each dataset to enable the model to quickly adapt to new bacterial detection tasks; the multi-task learning framework is used to simultaneously perform learning and training on multiple bacterial detection tasks, specifically, the performance of the model on three tasks of bacterial detection, bacterial counting and antibiotic resistance prediction is trained simultaneously to enhance the model's generalization ability for new data; online learning technology is applied to perform rapid detection response integrated training on the bacterial detection ensemble pre-model according to the adaptive generalization adjustment data. Specifically, the gradient boosting machine (GBM) is selected as the basic model, 100 trees are set, the depth of each tree is 6, the minimum number of leaf node samples is 10, and the learning rate is 0.1, so as to obtain the bacterial detection ensemble training model.

[0090] Step S4: Perform performance evaluation on the bacteria detection integrated training model to generate training model performance evaluation data; perform intelligent iterative optimization on the bacteria detection integrated training model based on the training model performance evaluation data to obtain a bacteria detection integrated model; perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

[0091] In an embodiment of the present invention, a cross-validation method, such as k-fold cross-validation, is used to evaluate the performance of an integrated training model for bacterial detection; performance evaluation indicators, such as accuracy, recall, F1 score, and area under the ROC curve (AUC), are used to generate training model performance evaluation data; based on the performance evaluation data, deficiencies in model performance are identified, such as low accuracy on specific bacterial categories; intelligent optimization algorithms, such as bacterial foraging optimization algorithms, are used to adjust model parameters, for example, if the detection accuracy is low, the depth of the model detection type is adjusted; the generalization ability of the model is enhanced; based on the model generalization enhancement measures, the performance of the model on new data is monitored in real time and adaptively adjusted; a meta-learning strategy is used to enable the model to quickly adapt to new bacterial samples; an online learning algorithm is applied to enable the model to update its parameters in real time to adapt to new bacterial samples; and a model distillation technique is used to compress the knowledge of multiple pre-trained bacterial detection models into a smaller model to improve the reasoning speed and efficiency of the model.

[0092] Preferably, step S2 comprises the following steps:

[0093] Step S21: applying multi-parameter statistics to the bacterial correction data to obtain bacterial statistical data; extracting bacterial biomarkers from the bacterial statistical data to obtain bacterial biomarker data; and identifying bacterial characteristics of the pre-classified bacterial data based on the bacterial biomarker data to obtain bacterial characteristic data;

[0094] Step S22: grouping the bacterial characteristic data to obtain bacterial characteristic grouping data; identifying bacterial groups according to the bacterial characteristic grouping data to obtain bacterial group data; dividing the bacterial group data into performance behaviors to generate bacterial behavior division data; determining the behavior pattern of the bacterial characteristic data according to the bacterial behavior division data to obtain bacterial behavior pattern data;

[0095] Step S23: extracting multi-mode standard rules from the bacterial behavior pattern data to generate bacterial multi-mode standard rules; comparing the pattern rules of the pre-classified bacterial data one by one according to the bacterial multi-mode standard rules to obtain bacterial key feature data; identifying bacterial type features from the bacterial key feature data to generate type feature data;

[0096] Step S24: Associating bacteria data to be detected with bacteria types through type feature data to obtain bacteria type association data; quantifying the degree of association of the bacteria type association data to generate type association quantification data; assigning type labels to the bacteria data to be detected based on the type association quantification data to obtain labels of the data to be detected; performing type feature mapping on the labels of the data to be detected based on the type feature data to obtain bacteria type mapping data.

[0097] In an embodiment of the present invention, statistical software is used to apply multi-parameter statistics to bacterial correction data to calculate statistical quantities such as means and variances between different samples; chemical analysis methods, such as gas chromatography-mass spectrometry (GC-MS), are used to qualitatively and quantitatively analyze components such as fatty acids and proteins in bacterial samples; specifically, characteristic chemical components are identified as bacterial biomarkers based on information such as retention time and mass-to-charge ratio; the extracted biomarker data is compared and analyzed with pre-classified bacterial data, and pattern recognition algorithms, such as principal component analysis (PCA) and cluster analysis, are used to classify bacteria with similar biomarker characteristics into one category to obtain bacterial characteristic data. According to the similarity of bacterial characteristic data, cluster analysis methods such as K-means or hierarchical clustering are used to divide bacterial characteristic data into different groups; analyze the grouped bacterial characteristic data to identify bacterial groups with common characteristics; determine the boundaries of bacterial groups by calculating the similarity within the group and the dissimilarity between the groups; divide the behavior according to the growth characteristics, metabolic type and other behavioral characteristics of the bacterial group; determine the behavior pattern of bacteria by combining the behavior division data and characteristic data of bacteria. Analyze the bacterial behavior pattern data and extract the standard rules of bacterial behavior characteristics; compare the extracted standard rules with the pre-classified bacterial data to identify the key characteristics of bacteria that meet the rules; classify the identified key characteristics of bacteria to determine the type characteristics of bacteria; compare the bacterial data to be detected with the known type characteristic data to find the most similar type characteristics; calculate the similarity between the data to be detected and the known type characteristic data to quantify the degree of association; assign corresponding type labels to the bacterial data to be detected according to the similarity calculation results; map the type labels of the data to be detected with the type characteristic data to obtain bacterial type mapping data.

[0098] As an example of the present invention, refer to Figure 2 As shown, in this example, step S3 includes:

[0099] Step S31: performing uniform normalization on the bacteria correction data to obtain uniform bacteria correction data; performing mapping enhancement on the bacteria type mapping data to generate type mapping enhancement data; demarcating the bacteria to-be-detected sample groups based on the bacteria uniform correction data and the type mapping enhancement data to generate to-be-detected sample group data; performing type feature correspondence on the to-be-detected sample group data according to the type mapping enhancement data to generate type feature correspondence data; performing type feature label extraction on the type feature correspondence data to generate type feature label data;

[0100] Step S32: performing bacterial colony feature recognition on the bacterial uniformity correction data to obtain bacterial colony feature data; performing colony distribution detection on the bacterial colony feature data to obtain bacterial colony distribution data; performing colony type number counting on the type feature corresponding data through the bacterial colony distribution data to obtain colony type number data; performing colony number label extraction on the colony type number data to generate colony number label data;

[0101] Step S33: performing bacterial growth state analysis on the bacterial colony characteristic data to obtain bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data; constructing a bacterial detection integrated model for the bacteria to be detected data according to the type characteristic label data, the colony quantity label data and the colony state label data to obtain a bacterial detection integrated pre-model;

[0102] Step S34: enhancing the generalization capability of the bacteria detection integrated pre-model to obtain a model generalization enhancement measure; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain adaptive generalization adjustment data;

[0103] Step S35: Perform rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model.

[0104] In an embodiment of the present invention, a normalization algorithm, such as minimum-maximum normalization or Z-score normalization, is used to scale the data to a uniform range, such as [0,1], to eliminate the influence of different dimensions and obtain uniformly corrected bacterial data; the bacterial type mapping data is expanded through data enhancement techniques, such as rotation, flipping, scaling, etc., to generate type mapping enhanced data; the normalized bacterial correction data is combined with the enhanced type mapping data, and the bacterial data to be detected is divided into sample groups through cluster analysis or decision tree methods to generate sample group data to be detected; a pattern recognition algorithm, such as a support vector machine (SVM), is used to match and identify type features of the sample group data to be detected; based on the type feature matching, the most representative type feature label is extracted through feature extraction techniques, such as principal component analysis (PCA). Use image processing techniques, such as edge detection and morphological operations, to identify the morphological characteristics of bacterial colonies; use spatial analysis methods, such as the nearest neighbor index, to analyze the distribution of colonies; combine colony distribution data and type feature corresponding data, use counters to count the number of different colony types; generate corresponding colony quantity labels based on the number of colony types; use time series analysis methods, such as growth curve analysis, to detect the growth status of bacteria; extract the growth status label of the colony based on the growth status analysis results; integrate type feature label data, colony quantity label data and colony status label data, and use a machine learning framework, such as random forest. Specifically, set 100 decision trees; each tree The maximum depth of the proposed method is 10, and no pruning is performed. At each decision node, a feature subset is randomly selected to select the best splitting attribute. The bootstrap method is used to extract samples from the original training set to train each tree. Thus, an integrated pre-model for bacterial detection is constructed. The generalization ability of the integrated pre-model for bacterial detection is enhanced by regularization techniques, such as L1 or L2 regularization, to obtain a model generalization enhancement measure. According to the model generalization enhancement measure, the integrated pre-model for bacterial detection is adaptively generalized, and the model parameters are dynamically adjusted to achieve adaptive generalization. The data after the model adaptive generalization adjustment is used to perform rapid iterative training to obtain an integrated training model for bacterial detection.

[0105] As an example of the present invention, refer to Figure 3 As shown, in this example, step S33 includes:

[0106] Step S331: performing colony contour recognition on the bacterial colony characteristic data to generate bacterial colony contour data; performing morphological measurement on the bacterial colony contour data to obtain bacterial morphological data;

[0107] Step S332: performing a colony two-dimensional morphological image scanning on the bacterial colony characteristic data based on the bacterial morphological data to generate a colony two-dimensional morphological image; performing a colony three-dimensional morphological reconstruction on the colony two-dimensional morphological image to obtain three-dimensional colony morphological data;

[0108] Step S333: performing bacterial volume detection on the three-dimensional colony morphology data to obtain bacterial volume data; performing bacterial division identification on the three-dimensional colony morphology data to obtain bacterial division state data; determining bacterial growth state of the bacteria to be detected data according to the bacterial division state data and the bacterial volume data to generate bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data;

[0109] Step S334: using the type feature label data, colony quantity label data and colony status label data as a bacteria detection integrated model classifier component to obtain a model classifier component; inputting the to-be-detected sample group data into the model classifier component to obtain detection sample classification data;

[0110] Step S335: Perform detection integration and fusion on the group data of the samples to be detected according to the detection sample classification data to obtain a bacteria detection integrated pre-model.

[0111] In the embodiment of the present invention, image processing technology is used to identify the outline of the colony through threshold segmentation and edge detection algorithms (such as Canny algorithm) to generate bacterial colony outline data; morphological analysis is performed on the identified colony outline, including calculating the area, perimeter, shape index, etc. of the colony to obtain bacterial morphological data; scanning technology, such as confocal microscopy, is used to obtain a two-dimensional morphological image of the bacterial colony; a three-dimensional reconstruction algorithm, such as a two-photon microscope or optical coherence tomography technology, is used to reconstruct the three-dimensional morphological data of the colony from the two-dimensional image; the volume of the bacteria is determined by using the three-dimensional morphological data using a volume calculation formula; the three-dimensional morphological data is analyzed and the image is used to obtain the three-dimensional morphological data. Image processing technology is used to identify the pattern and stage of bacterial division to obtain bacterial division status data; the growth state of bacteria is determined by using a pattern recognition algorithm in combination with the division status and volume data to generate bacterial growth status data; based on the growth status data, colony status labels representing different growth stages are extracted to generate colony status label data; the extracted various label data are integrated into the machine learning model as feature input to obtain the model classifier component; the data of the sample to be tested is input into the classifier component to obtain the classification data of the test sample; the classified data is fused using an integrated learning method, such as a random forest or a gradient boosting machine, to obtain an integrated pre-model for bacterial detection.

[0112] Preferably, step S34 includes the following steps:

[0113] Step S341: determining the classification feature layer of the bacteria detection integrated pre-model to obtain classification feature layer information; performing feature structure recognition on the classification feature layer information to obtain classification feature structure data; performing feature structure slicing on the model classifier component according to the classification feature structure data to generate feature structure slicing data; performing type range segmentation on the type feature label data through the feature structure slicing data to obtain type segmentation data;

[0114] Step S342: performing colony magnitude gradient grouping on the colony quantity label data according to the type segmentation data to obtain colony magnitude gradient data; performing colony state cycle determination on the colony state label data according to the colony magnitude gradient data to obtain colony state cycle data;

[0115] Step S343: performing multi-dimensional feature fusion on the type subdivision data, colony magnitude gradient data and colony state cycle data to generate multi-dimensional feature fusion data; performing feature autoencoding compression on the multi-dimensional feature fusion data to obtain multi-dimensional feature compression data;

[0116] Step S344: performing feature multi-scale recognition on the multi-dimensional feature compression data to generate multi-scale feature data; performing detection label weight extraction on the model classifier component to obtain detection weight data; performing feature weight mapping on the detection weight data through the multi-scale feature data to generate multi-scale detection weight data; performing dynamic weight generalization enhancement on the bacteria detection integrated pre-model through the multi-scale detection weight data to obtain model generalization enhancement measures;

[0117] Step S345: adjusting the detection priority of the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain priority adjustment data; and performing adaptive learning rate optimization on the priority adjustment data to obtain adaptive generalization adjustment data.

[0118] In an embodiment of the present invention, a machine learning algorithm, such as a random forest, is used to evaluate the feature importance of the model to determine the feature layer information that contributes most to the classification; a feature selection method, such as recursive feature elimination (RFE), is used to identify the most representative feature structure to obtain classification feature structure data; a feature extraction technique, such as principal component analysis (PCA), is used to perform dimensionality reduction processing on the classification feature structure data to generate feature structure slice data; a clustering algorithm, such as K-means, is used to segment the type feature label data according to the feature structure slice data to obtain type segmentation data. According to the frequency of type features in the type segmentation data, the quantile method is used to divide the colony quantity label data into different magnitude gradients to obtain colony magnitude gradient data; the colony magnitude gradient data and time series data are combined to determine the periodic changes in the colony state through a periodic analysis method to obtain colony state periodic data; the type segmentation data, colony magnitude gradient data and colony state periodic data are combined to perform multi-dimensional feature fusion to generate multi-dimensional feature fusion data; the autoencoder is used to compress the multi-dimensional feature fusion data to extract more representative features to obtain multi-dimensional feature compression data; multi-scale analysis methods, such as wavelet transform, are applied to the multi-dimensional feature compression data to identify features at different scales to generate multi-scale feature data; In the model classifier component, the weight of each detection label is determined by the gradient ascent method in the training process to obtain the detection weight data; the multi-scale feature data is combined with the detection weight data, and the feature weight mapping is performed by the weighted average method to generate multi-scale detection weight data; according to the multi-scale detection weight data, the weight of the model is dynamically adjusted to enhance the generalization ability of the model for unseen samples, and the model generalization enhancement measures are obtained; according to the model generalization enhancement measures, the detection priority of different features is adjusted to reflect their contribution to the final classification decision, and the priority adjustment data is obtained; using an adaptive learning rate algorithm, such as the Adam optimizer, the learning rate of the model is dynamically adjusted according to the priority adjustment data to obtain adaptive generalization adjustment data.

[0119] Preferably, step S345 includes the following steps:

[0120] Step S3451: performing non-linear feature screening on the multi-scale detection weight data according to the model generalization enhancement measure to generate non-linear feature screening data; performing detection contribution calculation on the detection sample classification data according to the non-linear feature screening data to obtain detection contribution data;

[0121] Step S3452: performing contribution level division on the detection contribution data to obtain contribution level level data; performing detection weight sorting on the detection weight data based on the contribution level level data to generate weight sorting data; performing detection priority weighted calculation on the weight sorting data to obtain detection priority weighted data;

[0122] Step S3453: assigning multi-scale priority to the detection weight data according to the detection priority weighted data and the multi-scale feature data to obtain multi-scale priority assignment information; performing detection priority transformation on the multi-scale detection weight data based on the multi-scale priority assignment information to obtain priority adjustment data;

[0123] Step S3454: performing dynamic feature sensitivity detection on the priority adjustment data to obtain feature sensitivity data; performing sensitivity level classification on the weight sorting data according to the feature sensitivity data to obtain weight sensitivity level data; performing weight feature learning rate calculation on the weight sensitivity level data to generate feature learning rate data;

[0124] Step S3455: dynamically divide the feature learning rate data according to the detection priority weighted data to obtain learning rate division data; and adaptively optimize the priority adjustment data according to the dynamic learning rate division to obtain adaptive generalization adjustment data.

[0125] In an embodiment of the present invention, a nonlinear feature screening technology is used, such as using kernel principal component analysis (KernelPCA) or autoencoders, to perform feature conversion and dimensionality reduction on multi-scale detection weight data, and screen out nonlinear features that best represent the data structure; by calculating the degree of influence of each feature on the model prediction result, for example, using the feature importance score in the random forest or the gradient-based feature importance method, the contribution of the nonlinear feature screening data to the detection sample classification data is evaluated to generate detection contribution data; using hierarchical clustering or K-means clustering algorithm, the features are divided into different contribution levels according to the detection contribution data to obtain contribution level data; according to the contribution level data, the detection weight data is sorted to generate weight sorting data; using a weighted summation method, according to the feature weights in the weight sorting data and their corresponding contributions, the detection priority weighted data is calculated; combining the detection priority weighted data and the multi-scale feature data, multi-scale priority assignment is performed on each feature to generate multi-scale priority assignment information; according to the multi-scale priority assignment information, the multi-scale detection weight data is adjusted. The weight values ​​in the data are used to reflect the priorities of different features in the detection, and priority adjustment data is obtained; the gradient information of the model or the variance analysis of the features is used to perform feature sensitivity detection on the priority adjustment data to obtain feature sensitivity data; according to the feature sensitivity data, the threshold segmentation or similarity measurement method is used to divide the weight sorting data into sensitivity levels to obtain weight sensitivity level data; an adaptive learning rate algorithm is used, such as the learning rate scheduling strategy in AdaBoost or the gradient descent method, to calculate the learning rate of the feature according to the weight sensitivity level data to generate feature learning rate data; the learning rate of each feature is dynamically adjusted in combination with the detection priority weighted data and the feature learning rate data to optimize the convergence speed and classification performance of the model to obtain learning rate division data; according to the learning rate division data, the priority adjustment data is adaptively optimized in learning rate. Specifically, Adam is selected as the optimizer, and the initial learning rate is set to 0.001. During the training process, the Adam optimizer will automatically adjust the learning rate of each parameter according to the gradient information, thereby obtaining adaptive generalization adjustment data, and finally completing the dynamic adjustment of feature weights in the entire model training process.

[0126] Preferably, step S35 includes the following steps:

[0127] Step S351: performing single feature matching on the model classifier component according to the adaptive generalization adjustment data to generate single feature matching data; performing multi-scale feature association matching on the sample group data to be detected based on the single feature matching data to obtain multi-scale feature association data;

[0128] Step S352: performing feature detection interconnection on the multi-dimensional feature fusion data according to the multi-scale feature association data to obtain detection feature interconnection data; performing label parameter extraction on the detection feature interconnection data to generate interconnection label parameters;

[0129] Step S353: performing detection priority interconnection on the detection priority weighted data based on the interconnection tag parameters to obtain detection priority interconnection data; performing detection contribution increment on the detection priority interconnection data to obtain detection contribution increment data; performing detection weight re-ranking on the weight ranking data according to the detection contribution increment data to generate weight ranking response data;

[0130] Step S354: Perform a fast learning rate response on the learning rate division data according to the weight sorting response data to obtain the learning rate fast response data; perform a fast pre-detection on the sample group data to be detected through the learning rate fast response data to obtain the sample pre-detection data; input the sample pre-detection data into the bacteria detection integrated pre-model for fast detection response training to obtain the bacteria detection integrated training model.

[0131] In an embodiment of the present invention, the contribution of each feature to the model is determined by using the gradient information of the model and the feature importance score; the most influential features are screened out through a feature selection algorithm, such as recursive feature elimination (RFE), to generate single feature matching data; the single feature matching data is compared with feature data of different scales, and correlation analysis or pattern recognition technology, such as K-nearest neighbor (KNN) algorithm, is specifically used to identify and match multi-scale features to obtain multi-scale feature association data; feature fusion technology, such as a multi-scale feature fusion network, is used to combine multi-scale feature association data with multi-dimensional feature fusion data to achieve interconnection between features and obtain detection feature interconnection data; the detection feature interconnection data is analyzed through a supervised learning algorithm, such as a support vector machine (SVM), and key label parameters for classification decisions are extracted to generate interconnection label parameters; a decision tree algorithm is used to perform feature determination on the interconnection label parameters and the detection priority weighted data, specifically determining the priorities of different features to obtain detection priority interconnection data. Combine the detection priority interconnection data and the prediction results of the model, calculate the incremental contribution of each feature to the final decision, and obtain the incremental detection contribution data; re-evaluate the weight of each feature in the weight ranking data based on the incremental detection contribution data, and generate weight ranking response data; according to the weight ranking response data, use an adaptive learning rate algorithm, such as AdaBoost, to dynamically adjust the learning rate of each feature to obtain learning rate fast response data; use the learning rate fast response data to perform preliminary detection on the sample group data to be detected, quickly screen out bacterial samples, and obtain sample pre-detection data; input the sample pre-detection data into the bacterial detection integrated pre-model, and perform rapid iterative training to optimize the model parameters to obtain the bacterial detection integrated training model; according to the model generalization enhancement measures, adjust the detection priority in the detection integrated pre-model, optimize the decision path of the model, and obtain priority adjustment data; combine the priority adjustment data and feature learning rate data, use an adaptive learning rate algorithm, such as AdaGrad, to optimize the learning rate of the model, and obtain adaptive generalization adjustment data.

[0132] Preferably, step S4 comprises the following steps:

[0133] Step S41: Divide the to-be-detected sample group data into a test set and a validation set to obtain a to-be-detected sample test set and a to-be-detected sample validation set;

[0134] Step S42: input the test set of samples to be detected into the bacteria detection integrated training model to measure the detection accuracy and generate test set accuracy data; input the test set of samples to be detected into the bacteria detection integrated training model to record the recall rate and generate test set recall rate data; integrate the test set accuracy rate data and the test set recall rate data to obtain the test set performance data;

[0135] Step S43: input the verification set of samples to be tested into the bacteria detection integrated training model to measure the detection accuracy and generate verification set accuracy data; input the verification set of samples to be tested into the bacteria detection integrated training model to record the recall rate and generate verification set recall rate data; integrate the verification set accuracy rate data and the verification set recall rate data to obtain the verification set performance data;

[0136] Step S44: performing a performance comparison between the test set performance data and the validation set performance data to generate training model performance evaluation data; iteratively optimizing the bacteria detection integrated training model according to the training model performance evaluation data to obtain a bacteria detection integrated model;

[0137] Step S45: Perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

[0138] In an embodiment of the present invention, a data segmentation technology is used, such as simple random sampling or stratified sampling, to divide the data of the sample group to be detected into a test set of the sample to be detected and a verification set of the sample to be detected; specifically, the data set is divided according to a certain ratio, such as 70% as a training set, 15% as a verification set, and 15% as a test set; the test set is detected by using a bacteria detection integrated training model, and the accuracy is calculated by comparing the detection results with the actual labels to generate test set accuracy data; similarly, the test set is detected by using a bacteria detection integrated training model, the recall rate is calculated, and the test set recall rate data is generated; using data fusion technology, the test set accuracy data and the test set recall rate data are integrated to obtain the test set performance data. The same method is used to calculate the accuracy of the validation set to generate the validation set accuracy data; the recall rate of the validation set is recorded to generate the validation set recall rate data; the accuracy and recall rate data are integrated to obtain the validation set performance data; the performance data of the test set and the validation set are compared to evaluate the performance of the model on different data sets and generate the training model performance evaluation data; according to the performance evaluation data, the model parameters are adjusted. Specifically, if the model shows overfitting (the training set performs well and the test set performs poorly), the overfitting is reduced by adjusting the learning rate or adding dropout and other techniques; if the model is underfitting (both the training set and the test set perform poorly), the model complexity is increased or more training data is provided to obtain a bacterial detection integrated model; the bacterial detection data to be detected is subjected to bacterial detection based on the bacterial detection integrated model to obtain a bacterial detection report.

[0139] Preferably, step S44 includes the following steps:

[0140] Step S441: perform similarity measurement on the test set performance data and the validation set performance data to generate performance similarity data; perform similarity deviation calculation on the performance similarity data to obtain performance similarity deviation data; perform detection response time difference calculation on the test set performance data and the validation set performance data to obtain response time difference data;

[0141] Step S442: performing performance evaluation on the integrated training model for bacteria detection through performance similarity deviation data and response time difference data, and generating training model performance evaluation data;

[0142] Step S443: performing anomaly detection feedback on the bacteria detection integrated training model according to the training model performance evaluation data to obtain anomaly detection feedback data; performing an association parameter improvement operation on the multi-scale feature association data according to the anomaly detection feedback data to generate association parameter improvement data;

[0143] Step S444: performing error back propagation on the performance similarity deviation data according to the associated parameter improvement data to obtain error back propagation data; performing loss function back propagation on the detection priority weighted data according to the error back propagation data to obtain loss function back propagation data; performing regularization constraints on the bacteria detection integrated training model according to the loss function back propagation data to obtain regularization constraint data;

[0144] Step S445: performing multi-task collaborative training on the sample pre-detection data according to the regularized constraint data to obtain multi-task collaborative data; iteratively optimizing the bacteria detection integrated training model according to the multi-task collaborative data to obtain the bacteria detection integrated model.

[0145] In an embodiment of the present invention, a metric method such as cosine similarity or Euclidean distance is used to calculate the similarity between the performance data of the test set and the performance data of the validation set; the concept of deviation in statistics is used to calculate the absolute deviation and relative deviation of the performance similarity; by comparing the response time of the test set and the validation set, the difference between the two is calculated to obtain the response time difference data; combining the performance similarity deviation data and the response time difference data, specifically through a comprehensive evaluation method, such as a weighted average method, to generate training model performance evaluation data; based on the performance evaluation data, anomaly detection is performed using anomaly detection algorithms, such as DBSCAN, and anomaly detection feedback data is generated; according to the anomaly detection feedback data, the association parameters of the multi-scale feature association data are adjusted to specifically improve the accuracy and robustness of the model; using the error back propagation algorithm, the data adjustment error is improved according to the association parameters, and back propagated to the model parameters to obtain the error back propagation data; the error back propagation data is applied to the loss function, and the gradient of the parameters is calculated by the chain rule to achieve The loss function is back-propagated to obtain the back-propagated data of the loss function; based on the back-propagated data of the loss function, a regularization term, such as L1 or L2 regularization, is introduced to perform regularization constraints on the integrated training model for bacteria detection to obtain regularized constrained data; according to the regularized constrained data, a multi-task learning strategy is adopted to perform multi-task collaborative training. Specifically, a neural network including shared layers and task-specific layers is constructed, wherein the shared layers (such as the first few convolutional layers) are responsible for extracting common features, such as edges and textures in bacterial images; each task has its own specific layer, and task-related features are further extracted on the basis of the shared layers. For example, for the task of bacteria classification, a deeper convolutional layer is required to identify the morphological characteristics of bacteria; and for the task of bacteria counting, a fully connected layer is required to process spatial information; finally, multi-task collaborative data is obtained; using the multi-task collaborative data, the integrated training model for bacteria detection is iteratively optimized, specifically, through multiple iterative trainings, the parameters of the shared layers and the task-specific layers are continuously updated to obtain the final integrated model for bacteria detection.

[0146] Preferably, step S444 includes the following steps:

[0147] Step S4441: determining the parameter improvement range of the associated parameter improvement data to obtain associated parameter range data; calculating the deviation degree of the performance similarity deviation data to obtain performance deviation degree data;

[0148] Step S4442: mapping the associated parameter range data with the performance deviation degree data to generate deviation parameter mapping data; performing error impact identification on the deviation parameter mapping data to obtain error impact data;

[0149] Step S4443: performing error back propagation on the multi-scale feature association data according to the error influence data to obtain error back propagation data; performing detection network layer extraction on the bacteria detection integrated training model to obtain detection network layer data; performing loss function calculation layer by layer on the detection network layer data according to the detection priority weighted data to obtain the partial derivative of the loss function;

[0150] Step S4444: The partial derivative of the loss function is transmitted back to the detection network layer to obtain the loss function back-propagation data; the regularization function is determined according to the anomaly detection feedback data to obtain the regularization function; the regularization function is added to the loss function back-propagation data to generate the back-propagation function information; the bacteria detection integrated training model is regularized and constrained according to the back-propagation function information to obtain the regularization constraint data.

[0151] In an embodiment of the present invention, by analyzing the changing trend and distribution of multi-scale feature association data, the adjustable range of the association parameter is specifically determined; by using a statistical analysis method, such as calculating the minimum value, maximum value and standard deviation of the parameter, the association parameter range data is obtained; by using a deviation analysis method in statistics, such as the root mean square deviation (RMSD), the deviation degree of the performance similarity deviation data is calculated to obtain performance deviation degree data; the association parameter range data is compared with the performance deviation degree data, and a mapping algorithm, such as nearest neighbor mapping or multivariate linear regression, is used to generate deviation parameter mapping data; using pattern recognition technology, such as support vector machine (SVM) or decision tree, the deviation parameter mapping data is analyzed to identify errors that have a significant impact on model performance to obtain error impact data; according to the error impact data, an error back propagation algorithm, such as the chain rule, is used to adjust the multi-scale feature association data to obtain error back propagation data; by analyzing the structure of the bacterial detection integrated training model, the key detection network layer is identified to obtain the detection network layer data; in combination with the detection priority weighted data, a loss function, such as cross entropy loss or mean square error loss, is applied to the detection network layer data to perform layer-by-layer Calculate and obtain the partial derivative of the loss function; use the gradient descent method to reverse the partial derivative of the loss function back to the detection network layer, update the network parameters, and obtain the loss function back-propagation data; based on the anomaly detection feedback, select a suitable regularization function; if the model shows signs of overfitting, select L1 regularization (Lasso) or L2 regularization (Ridge) to reduce the model complexity; L1 regularization promotes the sparsity of model weights by adding the L1 norm (sum of absolute values) of the weights to the loss function, which helps feature selection; L2 regularization adds the L2 norm (sum of squares) of the weights to the loss function. ) to limit the size of the model weights to avoid overfitting; thereby obtaining a regularization function; adding the regularization function to the back-propagation data of the loss function to generate back-propagation function information; applying the back-propagation function information to perform regularization constraints on the bacteria detection integrated training model, optimize the model performance, and obtain regularized constraint data; using the regularized constraint data, perform multi-task collaborative training on the sample pre-detection data, such as using a multi-task learning framework to obtain multi-task collaborative data; based on the multi-task collaborative data, iteratively optimize the bacteria detection integrated training model, adjust the model parameters, and obtain the bacteria detection integrated model.

[0152] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0153] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for constructing an integrated model for bacterial detection, characterized in that: The following steps are involved: Step S1: obtaining a bacteria original data set; performing data denoising on the bacteria original data set to obtain bacteria data to be detected; performing pre-classification on the bacteria data to be detected to obtain pre-classified bacteria data; performing data batch identification on the pre-classified bacteria data to obtain batch data to be detected; performing batch deviation detection on the batch data to be detected to obtain bacteria batch deviation data; performing batch correction on the bacteria data to be detected according to the bacteria batch deviation data to obtain bacteria correction data; Step S2: performing bacterial characteristic recognition on the pre-classified bacterial data according to the bacterial correction data to obtain bacterial characteristic data; performing bacterial behavior pattern determination on the bacterial characteristic data to obtain bacterial behavior pattern data; performing bacterial type characteristic recognition on the pre-classified bacterial data according to the bacterial behavior pattern data to generate type characteristic data; performing type characteristic mapping on the bacterial data to be detected based on the type characteristic data to obtain bacterial type mapping data; Step S3: constructing a bacteria detection integrated model for bacteria data to be detected based on bacteria correction data and bacteria type mapping data to obtain a bacteria detection integrated pre-model; enhancing the generalization capability of the bacteria detection integrated pre-model to obtain model generalization enhancement measures; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measures to obtain adaptive generalization adjustment data; performing rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model; Step S4: Perform performance evaluation on the bacteria detection integrated training model to generate training model performance evaluation data; perform intelligent iterative optimization on the bacteria detection integrated training model based on the training model performance evaluation data to obtain a bacteria detection integrated model; perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

2. The method for constructing an integrated model for bacteria detection according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: applying multi-parameter statistics to the bacterial correction data to obtain bacterial statistical data; extracting bacterial biomarkers from the bacterial statistical data to obtain bacterial biomarker data; and identifying bacterial characteristics of the pre-classified bacterial data based on the bacterial biomarker data to obtain bacterial characteristic data; Step S22: grouping the bacterial characteristic data to obtain bacterial characteristic grouping data; identifying bacterial groups according to the bacterial characteristic grouping data to obtain bacterial group data; dividing the bacterial group data into performance behaviors to generate bacterial behavior division data; determining the behavior pattern of the bacterial characteristic data according to the bacterial behavior division data to obtain bacterial behavior pattern data; Step S23: extracting multi-mode standard rules from the bacterial behavior pattern data to generate bacterial multi-mode standard rules; comparing the pattern rules of the pre-classified bacterial data one by one according to the bacterial multi-mode standard rules to obtain bacterial key feature data; identifying bacterial type features from the bacterial key feature data to generate type feature data; Step S24: Associating bacteria data to be detected with bacteria types through type feature data to obtain bacteria type association data; quantifying the degree of association of the bacteria type association data to generate type association quantification data; assigning type labels to the bacteria data to be detected based on the type association quantification data to obtain labels of the data to be detected; performing type feature mapping on the labels of the data to be detected based on the type feature data to obtain bacteria type mapping data.

3. The method for constructing an integrated model for bacteria detection according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing uniform normalization on the bacteria correction data to obtain uniform bacteria correction data; performing mapping enhancement on the bacteria type mapping data to generate type mapping enhancement data; demarcating the bacteria to-be-detected sample groups based on the bacteria uniform correction data and the type mapping enhancement data to generate to-be-detected sample group data; performing type feature correspondence on the to-be-detected sample group data according to the type mapping enhancement data to generate type feature correspondence data; performing type feature label extraction on the type feature correspondence data to generate type feature label data; Step S32: performing bacterial colony feature recognition on the bacterial uniformity correction data to obtain bacterial colony feature data; performing colony distribution detection on the bacterial colony feature data to obtain bacterial colony distribution data; performing colony type number counting on the type feature corresponding data through the bacterial colony distribution data to obtain colony type number data; performing colony number label extraction on the colony type number data to generate colony number label data; Step S33: performing bacterial growth state analysis on the bacterial colony characteristic data to obtain bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data; constructing a bacterial detection integrated model for the bacteria to be detected data according to the type characteristic label data, the colony quantity label data and the colony state label data to obtain a bacterial detection integrated pre-model; Step S34: enhancing the generalization capability of the bacteria detection integrated pre-model to obtain a model generalization enhancement measure; adaptively generalizing and adjusting the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain adaptive generalization adjustment data; Step S35: Perform rapid detection response integrated training on the bacteria detection integrated pre-model according to the adaptive generalization adjustment data to obtain a bacteria detection integrated training model.

4. The method for constructing an integrated model for bacteria detection according to claim 3, characterized in that: Step S33 includes the following steps: Step S331: performing colony contour recognition on the bacterial colony characteristic data to generate bacterial colony contour data; performing morphological measurement on the bacterial colony contour data to obtain bacterial morphological data; Step S332: performing a colony two-dimensional morphological image scanning on the bacterial colony characteristic data based on the bacterial morphological data to generate a colony two-dimensional morphological image; performing a colony three-dimensional morphological reconstruction on the colony two-dimensional morphological image to obtain three-dimensional colony morphological data; Step S333: performing bacterial volume detection on the three-dimensional colony morphology data to obtain bacterial volume data; performing bacterial division identification on the three-dimensional colony morphology data to obtain bacterial division state data; determining bacterial growth state of the bacteria to be detected data according to the bacterial division state data and the bacterial volume data to generate bacterial growth state data; performing colony state label extraction on the bacterial growth state data to generate colony state label data; Step S334: using the type feature label data, colony quantity label data and colony status label data as a bacteria detection integrated model classifier component to obtain a model classifier component; inputting the to-be-detected sample group data into the model classifier component to obtain detection sample classification data; Step S335: Perform detection integration and fusion on the group data of the samples to be detected according to the detection sample classification data to obtain a bacteria detection integrated pre-model.

5. The method for constructing an integrated model for bacteria detection according to claim 3, characterized in that: Step S34 includes the following steps: Step S341: determining the classification feature layer of the bacteria detection integrated pre-model to obtain classification feature layer information; performing feature structure recognition on the classification feature layer information to obtain classification feature structure data; performing feature structure slicing on the model classifier component according to the classification feature structure data to generate feature structure slicing data; performing type range segmentation on the type feature label data through the feature structure slicing data to obtain type segmentation data; Step S342: performing colony magnitude gradient grouping on the colony quantity label data according to the type segmentation data to obtain colony magnitude gradient data; performing colony state cycle determination on the colony state label data according to the colony magnitude gradient data to obtain colony state cycle data; Step S343: performing multi-dimensional feature fusion on the type subdivision data, colony magnitude gradient data and colony state cycle data to generate multi-dimensional feature fusion data; performing feature autoencoding compression on the multi-dimensional feature fusion data to obtain multi-dimensional feature compression data; Step S344: performing feature multi-scale recognition on the multi-dimensional feature compression data to generate multi-scale feature data; performing detection label weight extraction on the model classifier component to obtain detection weight data; performing feature weight mapping on the detection weight data through the multi-scale feature data to generate multi-scale detection weight data; performing dynamic weight generalization enhancement on the bacteria detection integrated pre-model through the multi-scale detection weight data to obtain model generalization enhancement measures; Step S345: adjusting the detection priority of the bacteria detection integrated pre-model according to the model generalization enhancement measure to obtain priority adjustment data; and performing adaptive learning rate optimization on the priority adjustment data to obtain adaptive generalization adjustment data.

6. The method for constructing an integrated model for bacteria detection according to claim 5, characterized in that: Step S345 includes the following steps: Step S3451: performing non-linear feature screening on the multi-scale detection weight data according to the model generalization enhancement measure to generate non-linear feature screening data; performing detection contribution calculation on the detection sample classification data according to the non-linear feature screening data to obtain detection contribution data; Step S3452: performing contribution level division on the detection contribution data to obtain contribution level level data; performing detection weight sorting on the detection weight data based on the contribution level level data to generate weight sorting data; performing detection priority weighted calculation on the weight sorting data to obtain detection priority weighted data; Step S3453: assigning multi-scale priority to the detection weight data according to the detection priority weighted data and the multi-scale feature data to obtain multi-scale priority assignment information; performing detection priority transformation on the multi-scale detection weight data based on the multi-scale priority assignment information to obtain priority adjustment data; Step S3454: performing dynamic feature sensitivity detection on the priority adjustment data to obtain feature sensitivity data; performing sensitivity level classification on the weight sorting data according to the feature sensitivity data to obtain weight sensitivity level data; performing weight feature learning rate calculation on the weight sensitivity level data to generate feature learning rate data; Step S3455: dynamically divide the feature learning rate data according to the detection priority weighted data to obtain learning rate division data; and adaptively optimize the priority adjustment data according to the dynamic learning rate division to obtain adaptive generalization adjustment data.

7. The method for constructing an integrated model for bacteria detection according to claim 3, characterized in that: Step S35 includes the following steps: Step S351: performing single feature matching on the model classifier component according to the adaptive generalization adjustment data to generate single feature matching data; performing multi-scale feature association matching on the sample group data to be detected based on the single feature matching data to obtain multi-scale feature association data; Step S352: performing feature detection interconnection on the multi-dimensional feature fusion data according to the multi-scale feature association data to obtain detection feature interconnection data; performing label parameter extraction on the detection feature interconnection data to generate interconnection label parameters; Step S353: performing detection priority interconnection on the detection priority weighted data based on the interconnection tag parameters to obtain detection priority interconnection data; performing detection contribution increment on the detection priority interconnection data to obtain detection contribution increment data; performing detection weight re-ranking on the weight ranking data according to the detection contribution increment data to generate weight ranking response data; Step S354: Perform a fast learning rate response on the learning rate division data according to the weight sorting response data to obtain the learning rate fast response data; perform a fast pre-detection on the sample group data to be detected through the learning rate fast response data to obtain the sample pre-detection data; input the sample pre-detection data into the bacteria detection integrated pre-model for fast detection response training to obtain the bacteria detection integrated training model.

8. The method for constructing an integrated model for bacteria detection according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Divide the to-be-detected sample group data into a test set and a validation set to obtain a to-be-detected sample test set and a to-be-detected sample validation set; Step S42: input the test set of samples to be detected into the bacteria detection integrated training model to measure the detection accuracy and generate test set accuracy data; input the test set of samples to be detected into the bacteria detection integrated training model to record the recall rate and generate test set recall rate data; integrate the test set accuracy rate data and the test set recall rate data to obtain the test set performance data; Step S43: input the verification set of samples to be tested into the bacteria detection integrated training model to measure the detection accuracy and generate verification set accuracy data; input the verification set of samples to be tested into the bacteria detection integrated training model to record the recall rate and generate verification set recall rate data; integrate the verification set accuracy rate data and the verification set recall rate data to obtain the verification set performance data; Step S44: performing a performance comparison between the test set performance data and the validation set performance data to generate training model performance evaluation data; iteratively optimizing the bacteria detection integrated training model according to the training model performance evaluation data to obtain a bacteria detection integrated model; Step S45: Perform bacteria detection on the bacteria data to be detected based on the bacteria detection integrated model to obtain a bacteria detection report.

9. The method for constructing an integrated model for bacteria detection according to claim 8, characterized in that: Step S44 includes the following steps: Step S441: perform similarity measurement on the test set performance data and the validation set performance data to generate performance similarity data; perform similarity deviation calculation on the performance similarity data to obtain performance similarity deviation data; perform detection response time difference calculation on the test set performance data and the validation set performance data to obtain response time difference data; Step S442: performing performance evaluation on the integrated training model for bacteria detection through performance similarity deviation data and response time difference data, and generating training model performance evaluation data; Step S443: performing anomaly detection feedback on the bacteria detection integrated training model according to the training model performance evaluation data to obtain anomaly detection feedback data; performing an association parameter improvement operation on the multi-scale feature association data according to the anomaly detection feedback data to generate association parameter improvement data; Step S444: performing error back propagation on the performance similarity deviation data according to the associated parameter improvement data to obtain error back propagation data; performing loss function back propagation on the detection priority weighted data according to the error back propagation data to obtain loss function back propagation data; performing regularization constraints on the bacteria detection integrated training model according to the loss function back propagation data to obtain regularization constraint data; Step S445: performing multi-task collaborative training on the sample pre-detection data according to the regularized constraint data to obtain multi-task collaborative data; iteratively optimizing the bacteria detection integrated training model according to the multi-task collaborative data to obtain the bacteria detection integrated model.

10. The method for constructing an integrated model for bacteria detection according to claim 9, characterized in that: Step S444 includes the following steps: Step S4441: determining the parameter improvement range of the associated parameter improvement data to obtain associated parameter range data; calculating the deviation degree of the performance similarity deviation data to obtain performance deviation degree data; Step S4442: mapping the associated parameter range data with the performance deviation degree data to generate deviation parameter mapping data; performing error impact identification on the deviation parameter mapping data to obtain error impact data; Step S4443: performing error back propagation on the multi-scale feature association data according to the error influence data to obtain error back propagation data; performing detection network layer extraction on the bacteria detection integrated training model to obtain detection network layer data; performing loss function calculation layer by layer on the detection network layer data according to the detection priority weighted data to obtain the partial derivative of the loss function; Step S4444: The partial derivative of the loss function is transmitted back to the detection network layer to obtain the loss function back-propagation data; the regularization function is determined according to the anomaly detection feedback data to obtain the regularization function; the regularization function is added to the loss function back-propagation data to generate the back-propagation function information; the bacteria detection integrated training model is regularized and constrained according to the back-propagation function information to obtain the regularization constraint data.

Images