Crowdsourcing rainfall data quality control method

Abstract

The embodiment of the invention discloses a crowdsourcing rainfall data quality control method, which comprises the following steps of S1, extracting multiple rainfall data statistical characteristics of rainfall observation points to establish a crowdsourcing rainfall data set with mathematical attributes; s2, training and preprocessing each rainfall data statistical feature in the crowdsourcing rainfall data set based on a machine learning algorithm to obtain a crowdsourcing rainfall data quality control model through training; and S3, carrying out importance sorting on the performance influence factors of the crowdsourcing rainfall data quality control model so as to calibrate the crowdsourcing rainfall data quality control model. The crowdsourcing rainfall data quality control model obtained by training and calibration solves the problems of subjectivity of previous crowdsourcing data quality judgment and incapability of automatically processing large-scale data, and provides key technical support for acquisition and application of hydrological crowdsourcing data, data governance of intelligent water conservancy and the like.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a method for quality control of crowdsourced rainfall data. Background Art

[0002] Precipitation is the most active link in the water cycle and a key input to the basin hydrological system. Its temporal and spatial accuracy directly affects the accuracy of hydrological simulation results. The abnormal temporal and spatial distribution of precipitation is also an important factor in causing natural disasters such as floods and landslides. Therefore, effective and accurate observation of the temporal and spatial distribution characteristics of rainfall plays an extremely important role in early warning and decision-making in hydrology, meteorology, agriculture, etc. The main ways to obtain traditional rainfall data include station, radar and satellite observations, etc., and their spatial resolution and temporal resolution have their own characteristics, but most of them are not suitable for small and medium-scale hydrological simulations, which require high temporal and spatial rainfall field data. The acquisition of high temporal and spatial rainfall fields can effectively improve the accuracy and forecast period of flood process simulation, increase the power generation of hydropower stations, and generate significant economic benefits. For urban construction management, the acquisition of high temporal and spatial rainfall fields can meet the needs of fine meteorological services in urbanized areas and improve the modernization level of urban disaster prevention, mitigation and emergency management.

[0003] In the era of big data, with the rapid development of the Internet of Things, non-traditional and diversified data acquisition technologies such as data crowdsourcing have promoted the generation of massive data. Different from professional data collection methods, data crowdsourcing mainly involves the public in the data collection process. Currently, data crowdsourcing has been increasingly studied worldwide, and crowdsourced rainfall data has the potential to make up for the shortcomings of traditional monitoring methods. Therefore, how to apply extensive crowdsourcing data in the era of big data to promote the improvement of the spatiotemporal resolution of rainfall observation is the most important and critical issue. At the same time, the quality control of crowdsourcing data has become the key to determine the further application of crowdsourcing science. Summary of the invention

[0004] Based on this, it is necessary to provide a crowdsourced rainfall data quality control method to solve the subjectivity of previous crowdsourced data quality judgments and the problem of being unable to automatically process large-scale data, providing key technical support for the collection and application of hydrological crowdsourced data and data governance of smart water conservancy.

[0005] A crowdsourced rainfall data quality control method comprises the following steps:

[0006] S1. Extract multiple rainfall data statistical features of rainfall observation points to establish a crowdsourced rainfall dataset with mathematical properties;

[0007] S2. Training and preprocessing the statistical features of various rainfall data in the crowdsourced rainfall data set based on a machine learning algorithm to train a crowdsourced rainfall data quality control model;

[0008] S3. rank the performance influencing factors of the crowd-sourced rainfall data quality control model in order to calibrate the crowd-sourced rainfall data quality control model.

[0009] In one embodiment, the step S1 of extracting a plurality of rainfall data statistical features comprises:

[0010] S11. Based on the surface rainfall fields with different coverages centered on the same preset rainfall observation point, the mathematical properties of rainfall in windows of different sizes are extracted with the preset rainfall observation point as the center, wherein the mathematical properties of rainfall include at least one of variance, maximum value, minimum value, range, mean value, standard deviation and absolute deviation;

[0011] S12. Based on the same preset rainfall observation point as the center, in the surface rainfall field within the coverage range of different radius distances, the rainfall mathematical correlation characteristics between the preset rainfall observation point and other rainfall observation points are extracted, and the rainfall mathematical correlation characteristics include at least one of the difference, the variation range and the gradient;

[0012] S13. Based on the rainfall fields formed at different times with the same preset rainfall observation point as the center, the correlation coefficient of two adjacent rainfall fields is extracted.

[0013] In one of the embodiments, the machine learning algorithm includes a supervised machine learning algorithm and an unsupervised machine learning algorithm;

[0014] The step S2 comprises:

[0015] S21. Based on the supervised machine learning algorithm, the error calibration method is used to randomly add noise values ​​to the rainfall observation points by introducing noise values ​​and calibrate them, and train and learn the statistical features of multiple rainfall data of the error rainfall observation points;

[0016] S22. Based on the unsupervised machine learning algorithm, the noise data is introduced into multiple rainfall data statistical features for automatic training and preprocessing.

[0017] In one embodiment, step S3 comprises:

[0018] S31. Using the SHAP interpretable algorithm to quantitatively interpret and rank the performance influencing factors of the crowdsourcing rainfall data quality control model;

[0019] S32. Using an extra tree algorithm to identify the best performance influencing factor combination of the crowdsourced rainfall data quality control model.

[0020] In one embodiment, in step S31, the SHAP value is calculated based on the following formula:

[0021]

[0022] Where f(x) is the model’s prediction for input x; x i is the i-th feature of input x; M is the number of features; φ0 is a constant representing the feature base value; φ i is the SHAP value of the i-th feature, representing the contribution of this feature to the prediction of f(x);

[0023]

[0024] Among them, S is the feature set, which represents the feature subset considered when the model predicts; |S| is the number of features in the set S; f x (S) is the prediction of the model when considering only the features in set S; f x (SU{i}) is the prediction of the model when considering the features in set S and feature i.

[0025] In one of the embodiments, in step S1, a part of the original rainfall data of the plurality of rainfall data statistical features is directly collected by a fixed sensor, and another part of the original rainfall data of the plurality of rainfall data statistical features is obtained by a preset processing after the data is directly collected by a movable sensor.

[0026] In one embodiment, the step of performing preset processing on the data directly collected by the movable sensor includes:

[0027] The movement path model associated with the movable sensor is coupled to the data quality control framework to simulate the dynamic movement path of the participants, and the deep coupling of the movement path model with the crowdsourced rainfall quality control model is realized. The control equations and parameters involved are as follows:

[0028]

[0029]

[0030] Among them, S i,t represents the position of participant i at time step t; L i,t represents the moving distance of participant i at time step t; θ i,t represents the moving direction of participant i at time step t; x and y are the coordinates of the rainfall observation point in space; α and φ represent the moving distance L i,t and the moving direction θ i,t The adjustment factor of i,t represents the Boolean movement state factor of participant i at time step t; the movement distance L i,t It is assumed to follow L i,t~U(0,L max ), U() means uniform distribution, L max represents the maximum distance that the participant can move within a time step; the moving direction θ i,t It is assumed to follow θ i,t ~U(0,2π), represents the movement probability of the rainfall observation point; and They represent the maximum and minimum moving probabilities of the rainfall observation points respectively; cx and cy represent the positions of the center of the rainfall observation point in the x and y directions respectively; r represents the radius of the center of the rainfall observation point; ζp represents the observation density gradient from the center of the rainfall observation point to the edge of the study area.

[0031] In one embodiment, it also includes:

[0032] S4. Using preset evaluation indicators to evaluate the prediction results of the crowdsourcing rainfall data quality control model.

[0033] In one embodiment, the evaluation indicators include WCFG and WCPJ, WCFG is used to describe the ability to capture rainfall variability at a small spatial and temporal scale, and WCPJ is used to describe the relative deviation of the total rainfall of rainfall events captured at the site scale within the duration and spatial coverage of rainfall;

[0034] The calculation formulas for WCFG and WCPJ are as follows:

[0035]

[0036] in, is the rainfall intensity estimated at time t based on crowdsourced observations; y i is the relevant ground truth rainfall intensity; i represents the sequence number of the rainfall observation point.

[0037] In one embodiment, it also includes:

[0038] S5. The ΔWCFG and ΔWCPJ evaluation indicators were used to evaluate the performance improvement of the crowdsourcing rainfall data quality control model after noise removal;

[0039] The calculation formulas for ΔWCFG and ΔWCPJ are as follows:

[0040]

[0041] Among them, WCFG before represents the WCFG value of the crowdsourced rainfall intensity before noise removal; WCFG after WCFG value representing the crowdsourced rainfall intensity after noise removal; WCPJ beforerepresents the WCPJ value of crowdsourced rainfall intensity before noise removal; WCPJ after Represents the WCPJ value of crowdsourced rainfall intensity after noise removal.

[0042] The crowdsourced rainfall data quality control method provided in the present application first extracts multiple rainfall data statistical features of rainfall observation points to establish a crowdsourced rainfall data set with mathematical properties, then trains and preprocesses the various rainfall data statistical features in the crowdsourced rainfall data set based on a machine learning algorithm to obtain a crowdsourced rainfall data quality control model through training, and finally calibrates the crowdsourced rainfall data quality control model by ranking the performance influencing factors of the crowdsourced rainfall data quality control model by importance. The crowdsourced rainfall data quality control model obtained through the above training and calibration solves the subjectivity of previous crowdsourced data quality judgments and the problem of being unable to automatically process large-scale data, and provides key technical support for the collection and application of hydrological crowdsourcing data, data governance of smart water conservancy, etc., and is of great significance for the promotion and application of big data, artificial intelligence, and Internet of Things technologies in the construction of smart water conservancy. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 is a flowchart of a crowd-sourced rainfall data quality control method in one embodiment;

[0045] Figure 2 is an interpolated surface rainfall field map in one embodiment;

[0046] Figure 3 A surface rainfall field map with noise observations identified in one embodiment;

[0047] Figure 4 is a distribution diagram of importance values ​​of the first 20 performance influencing factors of the crowdsourcing rainfall data quality model in one embodiment;

[0048] Figure 5 It is an interactive influence diagram of the change ratio of WCFG and the change ratio of WCPJ and the noise level and the noise amount of the crowdsourcing rainfall data quality model in one embodiment;

[0049] Figure 6 A distribution diagram of the change ratio of WCFG, the change ratio of WCPJ, and the AUC value of the crowdsourcing rainfall data quality control model trained under different scene conditions in an embodiment;

[0050] Figure 7 A distribution diagram of the change ratio of WCFG, the change ratio of WCPJ, and the AUC value of another crowdsourced rainfall data quality control model trained under different scene conditions in an embodiment;

[0051] Figure 8 ΔWCFG and ΔWCPJ value distribution diagram of a crowdsourced rainfall data quality control model trained under the same benchmark scenario based on different machine learning algorithms in one embodiment. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0054] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in this field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0055] like Figure 1 As shown, the present application provides a crowdsourced rainfall data quality control method, comprising the following steps:

[0056] S1. Extract multiple rainfall data statistical features of rainfall observation points to establish a crowdsourced rainfall dataset with mathematical properties;

[0057] S2. Based on a preset machine learning algorithm, the statistical features of various rainfall data in the crowdsourced rainfall data set are trained and preprocessed to obtain a crowdsourced rainfall data quality control model;

[0058] S3. Rank the performance influencing factors of the crowdsourcing rainfall data quality control model in order to calibrate the crowdsourcing rainfall data quality control model.

[0059] The crowdsourced rainfall data quality control method provided in the present application first extracts multiple rainfall data statistical features of rainfall observation points to establish a crowdsourced rainfall data set with mathematical properties, then trains and preprocesses the various rainfall data statistical features in the crowdsourced rainfall data set based on a machine learning algorithm to obtain a crowdsourced rainfall data quality control model through training, and finally calibrates the crowdsourced rainfall data quality control model by ranking the performance influencing factors of the crowdsourced rainfall data quality control model by importance. The crowdsourced rainfall data quality control model obtained through the above training and calibration solves the subjectivity of previous crowdsourced data quality judgments and the problem of being unable to automatically process large-scale data, and provides key technical support for the collection and application of hydrological crowdsourcing data, data governance of smart water conservancy, etc., and is of great significance for the promotion and application of big data, artificial intelligence, and Internet of Things technologies in the construction of smart water conservancy.

[0060] Specifically, establishing the crowdsourced rainfall data set with mathematical properties is to provide a data basis for subsequent training of a crowdsourced rainfall data quality control model. Optionally, step S1 of extracting multiple rainfall data statistical features includes:

[0061] S11. Based on the surface rainfall fields with different coverages centered on the same preset rainfall observation point, the mathematical properties of rainfall in windows of different sizes are extracted with the preset rainfall observation point as the center of the circle, and the mathematical properties of rainfall include at least one of variance, maximum value, minimum value, range, mean value, standard deviation and absolute deviation;

[0062] S12. Based on the same preset rainfall observation point as the center, in the surface rainfall field within the coverage range of different radius distances, the rainfall mathematical correlation characteristics between the preset rainfall observation point and other rainfall observation points are extracted, and the rainfall mathematical correlation characteristics include at least one of the difference, the variation range and the gradient;

[0063] S13. Based on the rainfall fields formed at different times with the same preset rainfall observation point as the center, the correlation coefficient of two adjacent rainfall fields is extracted.

[0064] Furthermore, in step S1, a part of the original rainfall data of the multiple rainfall data statistical features is directly collected by a fixed sensor (such as a fixed camera), and another part of the original rainfall data of the multiple rainfall data statistical features is directly collected by a movable sensor (such as an umbrella sensor) and obtained through preset processing.

[0065] The steps of pre-processing the data directly collected by the movable sensor include:

[0066] The movement path model associated with the movable sensor is coupled to the data quality control architecture to simulate the dynamic movement path of the participants and achieve deep coupling between the movement path model and the crowdsourcing rainfall quality control model. The control equations and parameters involved are as follows:

[0067]

[0068] Among them, S i,t represents the position of participant i at time step t; L i,t represents the moving distance of participant i at time step t; θ i,t represents the moving direction of participant i at time step t; x and y are the coordinates of the rainfall observation point in space; α and φ represent the moving distance L i,t and the moving direction θ i,t The initial values ​​of α and φ are 1 and 0 respectively. i,t+1 When outside the study area, the values ​​of α and φ are adjusted so that movement is limited to the boundaries of the study area; ω i,t represents the Boolean movement state factor of participant i at time step t; the movement distance L i,t It is assumed to follow L i,t ~U(0,L max ), U() means uniform distribution, L max represents the maximum distance that the participant can move within a time step; the moving direction θ i,t It is assumed to follow θ i,t ~U(0,2π), represents the movement probability of the rainfall observation point; and They represent the maximum and minimum moving probabilities of the rainfall observation points respectively; cx and cy represent the positions of the center of the rainfall observation point in the x and y directions respectively; r represents the radius of the center of the rainfall observation point; ζp represents the observation density gradient from the center of the rainfall observation point to the edge of the study area.

[0069] Specifically, as mentioned above, rainfall data may be directly collected by a fixed device (such as a fixed camera, which is immovable and has a fixed position), that is, a fixed sensor is arranged at a fixed rainfall observation point to collect rainfall data;

[0070] Rainfall data may also be data collected by devices such as umbrellas or cars equipped with movable sensors (the device is movable) after preset processing. In this case, the rainfall observation point is changing, and the device equipped with the movable sensor is also controlled by the participant. For example, the participant holds an umbrella equipped with an umbrella sensor and moves to realize the rainfall data collection of the umbrella sensor during the movement. For another example, the participant drives a car equipped with a car wiper sensor to realize the rainfall data collection of the car wiper sensor during the movement.

[0071] Furthermore, rainfall observations are usually point observations, so the calculation of surface rainfall must be done through interpolation. Specifically, for the methods of rainfall spatial interpolation, whether it is global interpolation (such as boundary interpolation, trend surface analysis) or local interpolation (such as Kriging method, Thiessen polygon), they are all based on the geographical location relationship between the interpolation point and the rainfall observation point, and the rainfall data are interpolated using the interpolation method.

[0072] Furthermore, the purpose of the machine learning algorithm is to label each crowdsourced rainfall observation as a noise observation or a regular observation. A noise observation is when the observed data shows an anomaly. For example, if a tipping bucket rain gauge is broken, it will report a zero value. When there is data from surrounding rainfall observations, this will be identified as an erroneous observation, that is, a noise observation.

[0073] In step S2, the statistical features of various rainfall data in the crowdsourced rainfall data set are trained through a machine learning algorithm. It can be understood that the statistical features of various rainfall data are taken as independent variables, and the crowdsourced rainfall observation results are taken as dependent variables. The complex relationship between the statistical features of rainfall data and the crowdsourced rainfall observation results is identified through a machine learning algorithm.

[0074] Figure 2 is an interpolated surface rainfall field map in an embodiment, Figure 3 is a surface rainfall field map for identifying noise observations in an embodiment, Figure 3 In the figure, the noise observation points are represented by black triangles. The machine learning algorithms include supervised machine learning algorithms and unsupervised machine learning algorithms. Specifically, the supervised machine learning algorithms include multi-layer neural networks and K nearest neighbor algorithms, and the unsupervised machine learning algorithms include isolation forests and K-means clustering algorithms. Step S2 includes:

[0075] S21. Based on the supervised machine learning algorithm, the error calibration method is used to randomly add noise values ​​to the rainfall observation points by introducing noise values ​​and calibrate them, and train and learn the statistical features of multiple rainfall data of the error rainfall observation points;

[0076] Specifically, step S21 refers to artificially introducing outliers into normal observations based on a supervised machine learning algorithm, and then marking them as abnormal observations for supervised training to learn the characteristics of such anomalies; that is, based on the feature mapping relationship of the predicted noise points (that is, the mapping relationship between the statistical characteristics of the rainfall data and the crowdsourced rainfall observation results), a supervised machine learning algorithm is used for training to identify abnormal features (such as a sudden 0 value (which may be a clogged rain gauge) or an extremely high abnormal value), and then the abnormal features are used to identify the noise points and then remove them.

[0077] S22. Based on the unsupervised machine learning algorithm, the noise data is introduced into multiple rainfall data statistical features for automatic training and preprocessing.

[0078] Specifically, step S22 refers to the introduction of abnormal values ​​into normal observations based on an unsupervised machine learning algorithm, and the automatic identification of noise data without calibration. In other words, unsupervised training is to sort the importance of various rainfall data statistical features without knowing the feature mapping relationship of the noise points in advance, and then select the features with higher importance, apply the model, and eliminate the noise points. The model here refers to the crowdsourcing rainfall data quality control model trained with historical data, which can be directly used to predict the noise observation of subsequent rainfall.

[0079] Specifically, the randomly segmented data can be used to train supervised learning algorithms and unsupervised learning algorithms, with 70% of the collected data used for training and 30% for testing, where the training set is used to train the model and the test set is used to verify the prediction accuracy of the model. The training uses the 5-Folds cross-validation method to avoid overfitting, and the input eigenvalues ​​are normalized to 0-1 by the min-max scaling method before training. The feature normalization step enables the crowdsourcing rainfall data quality control model obtained by training to exclude the influence of different rainfall element data ranges, making it transferable and comparable between different regions and scenarios, even if their original eigenvalues ​​are in different ranges.

[0080] Specifically, the formula involved in the minimum-maximum standard normalization method is as follows:

[0081]

[0082] Where X represents the value of the current data point; X norm Indicates the normalized current data value; X min Indicates the minimum value of the current data set; X max Indicates the maximum value of the current data set.

[0083] In addition, the 0 mean normalization method can be used for normalization, and the formula involved is as follows:

[0084]

[0085] Among them, μ and σ represent the mean and variance of the original training data set, respectively.

[0086] Optionally, data in the training data set can be enhanced. For example, data enhancement can be performed by cropping, adding noise, etc. Using data enhancement technology mainly adds small disturbances or changes to the training data. On the one hand, it can increase the training data, thereby improving the generalization ability of the model, and on the other hand, it can increase the noise data, thereby enhancing the robustness of the model.

[0087] In one embodiment, step S3 includes:

[0088] S31. Use the SHAP interpretable algorithm to quantitatively interpret and rank the performance influencing factors of the crowdsourced rainfall data quality control model;

[0089] Specifically, the contribution of each performance influencing factor of the crowdsourcing rainfall data quality control model to the prediction of the crowdsourcing rainfall data quality control model is calculated through the SHAP interpretable algorithm, so that we can understand how each performance influencing factor affects the prediction of the crowdsourcing rainfall data quality control model.

[0090] In step S31, the calculation of the SHAP value is based on the following formula:

[0091]

[0092] Where f(x) is the model’s prediction for input x; x i is the i-th feature of input x; M is the number of features; φ0 is a constant representing the feature base value; φ i is the SHAP value of the i-th feature, representing the contribution of this feature to the prediction of f(x);

[0093]

[0094] Among them, S is the feature set, which represents the feature subset considered when the model predicts; |S| is the number of features in the set S; f x (S) is the prediction of the model when considering only the features in set S; f x (SU{i}) is the prediction of the model considering the features in set S and feature i. Specifically, x in the SHAP value calculation formula i Refers to the statistical characteristics of rainfall data in the present invention, and the calculation result of this formula is Figure 4 The importance ranking of various statistical features of rainfall data.

[0095] S32. Use the Extra Trees algorithm to identify the best performance influencing factor combination for the crowdsourced rainfall data quality control model.

[0096] The extra tree algorithm is an ensemble learning method that constructs a decision tree and selects a feature subset for node splitting. Its principle can be summarized as randomly selecting m features from the feature set F, where m = |F|. For the selected feature f, a split point is randomly selected. f , divide the data set into two subsets D according to the split point left and D right ,in:

[0097] D left ={x∈D|x f ≤split f}

[0098] D right ={x∈D|x f >split f}

[0099] For the noise judgment problem involved in this case, the extra tree algorithm is obtained by majority voting of the prediction results of all trees.

[0100]

[0101] Where I is the indicator function, if Otherwise, it is 1, otherwise it is 0. Specifically, in the present invention, the feature set F represents multiple rainfall data statistical features, and the feature selection step iteratively deletes the least important performance influencing factors (measured by the feature importance in the additional tree) until the prediction performance of the crowdsourced rainfall data quality model decreases, and the best performance influencing factor combination is selected by accuracy.

[0102] By using the prediction accuracy value of the crowdsourcing rainfall data quality model as the metric feature of the crowdsourcing rainfall data quality model performance, according to the feature selection results, the number of performance influencing factors selected by the supervised learning algorithm is 20 ( Figure 4 ), while the number of performance influencing factors selected by the unsupervised learning algorithm is 5 ( Figure 4 The first 5 features in Figure 4The importance value distribution diagram of the top 20 performance influencing factors of the crowdsourced rainfall data quality control model in one embodiment is shown. The importance values ​​of the top 20 performance influencing factors of the crowdsourced rainfall data quality control model are calculated by the extra tree algorithm, and the top 20 performance influencing factors of the crowdsourced rainfall data quality control model are ranked in importance according to the calculation results of the importance values. Among all the candidate performance influencing factors, Iad,C and Iad,w_9 are the top two performance influencing factors that contribute the most. Both of these performance influencing factors are related to the difference in average rainfall conditions of target observations in a certain spatial area. This difference can fully represent the mutation caused by noise observations.

[0103] Table 1. Definition of various performance influencing factors of the crowdsourcing rainfall data quality model

[0104]

[0105]

[0106] Referring to Table 1, which shows the definition of each performance influencing factor of the crowdsourcing rainfall data quality model, the maximum and minimum values ​​of the window domain can be defined as the maximum rainfall intensity and the minimum rainfall intensity within the window range of different sizes, respectively;

[0107] The calculation formula for the standard deviation value of the window domain is as follows:

[0108]

[0109] Among them, σ represents the standard deviation, N represents the total number of values ​​in the data set, xi represents each value in the data set, and μ represents the mean of the data set.

[0110] The calculation formula for the average value of the window domain is as follows:

[0111] average=sum(Rn) / n

[0112] Among them, Rn represents the rainfall intensity at each rainfall observation point, and n represents the number of observation points.

[0113] It should be noted that I av,C (Absolute value) refers to the absolute value of the difference between the range and the sampling point within the preset coverage range, where the range is the range of the maximum rainfall intensity and the minimum rainfall intensity of each rainfall observation point within the preset coverage range, and the sampling point difference is the absolute value of the difference between the value of the monitoring point and the range. The formula involved is as follows:

[0114]

[0115] For example, within a certain range, the rainfall intensity values ​​of each rainfall observation point are (1, 3, 7, 5, 8, 10), and the value range is 10-1=9; for the point where the sampling point observation value is 3, I av,c =|3-9|=6, indicating whether the sample point value is close to the extreme value of the interval.

[0116] It should be noted that the last number in the mathematical symbol of the performance impact factor represents the size of the window range based on the window feature. For example, the performance impact factor I ad,w_9 , the number 9 indicates that the size of the window-based range is 9*9 spatial grids.

[0117] Optionally, the crowd-sourced rainfall data quality control method further includes:

[0118] S4. Use the preset evaluation indicators to evaluate the prediction results of the crowdsourcing rainfall data quality control model.

[0119] Specifically, the evaluation indicators include WCFG (root mean square error) and WCPJ (relative mean error). WCFG is used to describe the ability to capture rainfall variability at small time and space scales, and WCPJ is used to describe the relative deviation of the total rainfall of rainfall events captured at the site scale within the duration and spatial coverage of rainfall.

[0120] The calculation formulas for WCFG and WCPJ are as follows:

[0121]

[0122]

[0123] in, is the rainfall intensity estimated at time t based on crowdsourced observations; y i is the relevant ground truth rainfall intensity; i represents the sequence number of the rainfall observation point.

[0124] Furthermore, the crowdsourcing rainfall data quality control method further includes:

[0125] S5. The performance improvement of the crowdsourcing rainfall data quality control model after noise removal was evaluated using the ΔWCFG (change ratio of WCFG) and ΔWCPJ (change ratio of WCPJ) evaluation indicators;

[0126] The calculation formulas for ΔWCFG and ΔWCPJ are as follows:

[0127]

[0128] Among them, WCFG before represents the WCFG value of the crowdsourced rainfall intensity before noise removal; WCFG afterWCFG value representing the crowdsourced rainfall intensity after noise removal; WCPJ before represents the WCPJ value of crowdsourced rainfall intensity before noise removal; WCPJ after Represents the WCPJ value of crowdsourced rainfall intensity after noise removal.

[0129] In the analysis of the interactive effects of noise level and number of noisy observations, e.g. Figure 5 As shown in the figure, the change ratio (specifically the reduction ratio) of WCFG and the change ratio (specifically the reduction ratio) of WCPJ increase with the increase of noise level and noise amount. Specifically, the reduction ratio of WCFG and the reduction ratio of WCPJ are as high as 55.49% and 69.25%, respectively, and are achieved at higher noise level and noise amount levels. At the same time, there is no clear interaction between the influence of noise level and noise amount on the application of the crowdsourced rainfall data quality control method of the present invention. This observation can be partially explained by the fact that the crowdsourced rainfall data quality control method of the present invention has a relatively stable ability in identifying noisy crowdsourced observations.

[0130] The crowdsourced rainfall data quality control method of the present invention is transferable, which means that the crowdsourced rainfall data quality control model trained under the baseline scenario is directly applied to other regions or other noise level scenarios. The migration application of the crowdsourced rainfall data quality control model helps to solve the problem that regional data quality is low or data is scarce and cannot be retrained, while saving computing costs. In the migration application of the crowdsourced rainfall data quality control model, the crowdsourced rainfall data quality control model trained in one region is migrated and applied to regions with significantly different rainfall climate characteristics. The results of the migration application show that the crowdsourced rainfall data quality control method still plays a role in reducing rainfall field errors and noise observation identification.

[0131] Figure 6 The figure shows the change ratio of WCFG, the change ratio of WCPJ and the distribution of AUC values ​​of the crowdsourced rainfall data quality control model obtained by training under different scene conditions in one embodiment; it can be seen from the figure that the S1 benchmark scenario performs well in terms of both error reduction ratio and AUC, among which the average reduction of WCFG is close to 46.2%, the average reduction of WCFG is close to 48.6%, and the AUC judgment value is close to 0.7, showing a good recognition effect. In addition, the effect of migration in S2-S4 is slightly worse than that of local training application, and the variance of the effect of migration application in different regions is also slightly larger, but it can still reduce the error to a certain extent and ensure a certain recognition effect; Figure 71 is a distribution diagram of the change ratio of WCFG, the change ratio of WCPJ and the AUC value of another crowd-sourced rainfall data quality control model trained under different scenario conditions in an embodiment; wherein S1 represents the crowd-sourced rainfall data quality control model trained under the benchmark scenario (referred to as the benchmark crowd-sourced rainfall data quality control model); S2 represents the benchmark crowd-sourced rainfall data quality control model migrated and applied to region 1; S3 represents the benchmark crowd-sourced rainfall data quality control model migrated and applied to region 2; S4 represents the benchmark crowd-sourced rainfall data quality control model migrated and applied to region 3; It should be noted that Figure 6 and Figure 7 The crowdsourced rainfall data quality control model trained in the paper is based on different machine learning algorithms. Specifically, Figure 6 and Figure 7 The crowdsourcing rainfall data quality control models shown are trained based on kNN and MLPs algorithms respectively.

[0132] In machine learning, AUC is an important feature for evaluating the performance of classification models, especially in binary classification problems. The AUC value is a non-parametric estimate of the area under the ROC curve. The ROC curve is constructed by plotting the True Positive Rate (TPR) against the False Positive Rate (FPR) at different thresholds. The calculation method of AUC can be summarized as follows:

[0133]

[0134] Among them, TPR is the true positive rate, which is FPR is the false positive rate, specifically TP is a true positive example, FN is a false negative example, FP is a false positive example, and TN is a true negative example. Specifically, TP, FN, FP, and TN are the confusion matrices in the machine learning algorithm. In the present invention, TP (true positive example) means that it is a normal rainfall observation, and the result of the recognition judgment is also a normal observation; TN (true negative example) means that it is a noise observation, and the result of the recognition judgment is also a noise observation; FN (false negative example) means that it is a normal observation, but the result of the recognition judgment is a noise observation; FP (false positive example) means that it is a noise observation, but the result of the recognition judgment is a normal observation.

[0135] Taking into account the randomness of the crowdsourcing rainfall data quality control model simulation, the performance of the crowdsourcing rainfall data quality control model may be affected by the uncertainty caused by the random selection of crowdsourcing observations. The ΔWCFG and ΔWCPJ values ​​of the crowdsourcing rainfall data quality control model trained based on different machine learning algorithms in the same benchmark scenario are calculated to quantify the impact of uncertainty caused by the random selection of crowdsourcing observations. Specifically, the benchmark scenario is repeated 1000 times to simulate and analyze the impact of this uncertainty. This application uses a total of 4 machine learning algorithms, which are kNN, MLPs, iForest, and K-Means algorithms, where kNN refers to the K nearest neighbor algorithm, MLPs refers to the multi-level neural network algorithm, iForest refers to the isolated forest algorithm, and K-Means refers to the K-means clustering algorithm. Figure 8 As shown in the figure, the ΔWCFG and ΔWCPJ values ​​of the crowdsourcing rainfall data quality control model trained by each machine learning algorithm approximately form a normal distribution with a relatively small standard deviation, which also shows that the performance ranking of these machine learning algorithms is not affected by uncertainty.

[0136] It should be noted that, in addition to the four classic supervised / unsupervised machine learning algorithms (kNN, MLPs, iForest, and K-Means) used in the present invention, other algorithms such as support vector machines, random forests, and Gaussian process algorithms can also be used to train crowdsourced rainfall data quality control models.

[0137] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. All equivalent structural changes made using the contents of the present invention's specification and drawings, or directly / indirectly applied in other related technical fields, are included in the patent protection scope of the present invention.

Claims

1. A crowd-sourced rainfall data quality control method, characterized in that: The following steps are involved: S1. Extract multiple rainfall data statistical features of rainfall observation points to establish a crowdsourced rainfall dataset with mathematical properties; S2. Training and preprocessing the statistical features of various rainfall data in the crowdsourced rainfall data set based on a machine learning algorithm to train a crowdsourced rainfall data quality control model; S3. rank the performance influencing factors of the crowd-sourced rainfall data quality control model in order to calibrate the crowd-sourced rainfall data quality control model.

2. The crowdsourcing rainfall data quality control method according to claim 1, characterized in that: The step S1 of extracting statistical features of a plurality of rainfall data comprises: S11. Based on the surface rainfall fields with different coverages centered on the same preset rainfall observation point, the mathematical properties of rainfall in windows of different sizes are extracted with the preset rainfall observation point as the center, wherein the mathematical properties of rainfall include at least one of variance, maximum value, minimum value, range, mean value, standard deviation and absolute deviation; S12. Based on the same preset rainfall observation point as the center, in the surface rainfall field within the coverage range of different radius distances, the rainfall mathematical correlation characteristics between the preset rainfall observation point and other rainfall observation points are extracted, and the rainfall mathematical correlation characteristics include at least one of the difference, the variation range and the gradient; S13. Based on the rainfall fields formed at different times with the same preset rainfall observation point as the center, the correlation coefficient of two adjacent rainfall fields is extracted.

3. The crowdsourcing rainfall data quality control method according to claim 1, characterized in that: The machine learning algorithm includes a supervised machine learning algorithm and an unsupervised machine learning algorithm; The step S2 comprises: S21. Based on the supervised machine learning algorithm, the error calibration method is used to randomly add noise values ​​to the rainfall observation points by introducing noise values ​​and calibrate them, and train and learn the statistical features of multiple rainfall data of the error rainfall observation points; S22. Based on the unsupervised machine learning algorithm, the noise data is introduced into multiple rainfall data statistical features for automatic training and preprocessing.

4. The crowd-sourced rainfall data quality control method according to claim 1, characterized in that: The step S3 comprises: S31. Using the SHAP interpretable algorithm to quantitatively interpret and rank the performance influencing factors of the crowdsourcing rainfall data quality control model; S32. Using an extra tree algorithm to identify the best performance influencing factor combination of the crowdsourced rainfall data quality control model.

5. The crowd-sourced rainfall data quality control method according to claim 4, characterized in that: In step S31, the calculation of the SHAP value is based on the following formula: Where f(x) is the model’s prediction for input x; x i is the i-th feature of input x; M is the number of features; φ0 is a constant representing the feature base value; φ i is the SHAP value of the i-th feature, representing the contribution of this feature to the prediction of f(x); Among them, S is the feature set, which represents the feature subset considered when the model predicts; |S| is the number of features in the set S; f x (S) is the prediction of the model when considering only the features in set S; f x (SU{i}) is the prediction of the model when considering the features in set S and feature i.

6. The crowd-sourced rainfall data quality control method according to claim 1, characterized in that: In step S1, a part of the original rainfall data of the plurality of rainfall data statistical features is directly collected by a fixed sensor, and another part of the original rainfall data of the plurality of rainfall data statistical features is obtained by directly collecting data from a movable sensor after a preset processing.

7. The crowd-sourced rainfall data quality control method according to claim 6, characterized in that: The step of performing preset processing on the data directly collected by the movable sensor comprises: The movement path model associated with the movable sensor is coupled to the data quality control framework to simulate the dynamic movement path of the participants, and the deep coupling of the movement path model with the crowdsourced rainfall quality control model is realized. The control equations and parameters involved are as follows: Among them, S i,t represents the position of participant i at time step t; L i,t represents the moving distance of participant i at time step t; θ i,t represents the moving direction of participant i at time step t; x and y are the coordinates of the rainfall observation point in space; α and φ represent the moving distance L i,t and the moving direction θ i,t The adjustment factor of i,t represents the Boolean movement state factor of participant i at time step t; the movement distance L i,t It is assumed to follow L i,t ~U(0,L max ), U() means uniform distribution, L max represents the maximum distance that the participant can move within a time step; the moving direction θ i,t It is assumed to follow θ i,t ~U(0,2π), represents the movement probability of the rainfall observation point; and They represent the maximum and minimum moving probabilities of the rainfall observation points respectively; cx and cy represent the positions of the center of the rainfall observation point in the x and y directions respectively; r represents the radius of the center of the rainfall observation point; ζp represents the observation density gradient from the center of the rainfall observation point to the edge of the study area.

8. The crowd-sourced rainfall data quality control method according to claim 1, characterized in that: Also includes: S4. Using preset evaluation indicators to evaluate the prediction results of the crowdsourcing rainfall data quality control model.

9. The crowd-sourced rainfall data quality control method according to claim 8, characterized in that: The evaluation indicators include WCFG and WCPJ. WCFG is used to describe the ability to capture rainfall variability at small time and space scales, and WCPJ is used to describe the relative deviation of the total rainfall of rainfall events captured at the site scale within the duration and spatial coverage of rainfall. The calculation formulas for WCFG and WCPJ are as follows: in, is the rainfall intensity estimated at time t based on crowdsourced observations; y i is the relevant ground truth rainfall intensity; i represents the sequence number of the rainfall observation point.

10. The crowd-sourced rainfall data quality control method according to claim 9, characterized in that: Also includes: S5. The ΔWCFG and ΔWCPJ evaluation indicators were used to evaluate the performance improvement of the crowdsourcing rainfall data quality control model after noise removal; The calculation formulas for ΔWCFG and ΔWCPJ are as follows: Among them, WCFG before represents the WCFG value of the crowdsourced rainfall intensity before noise removal; WCFG after WCFG value representing the crowdsourced rainfall intensity after noise removal; WCPJ before represents the WCPJ value of crowdsourced rainfall intensity before noise removal; WCPJ after Represents the WCPJ value of crowdsourced rainfall intensity after noise removal.

Images