Information dissemination risk prediction method and device, storage medium and electronic device

Abstract

The present application provides a method and device for predicting information dissemination risk, a storage medium and an electronic device, the method comprising: obtaining the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point based on the comment data generated by the manuscript to be predicted in each publishing channel according to a preset cycle, drawing a trend chart of the total number of comments and a trend chart of the number of negative comments, and calculating the proportion of negative comments, and determining the target alarm threshold based on the trend chart of the total number of comments and the trend chart of the number of negative comments. If the proportion of negative comments is not greater than the target alarm threshold, it is determined that the manuscript to be predicted does not have an information dissemination risk, otherwise, it is determined that the manuscript to be predicted has an information dissemination risk. It can be seen that the present application scheme realizes the prediction of information dissemination risk, provides more response time for handling risks, thereby reducing information dissemination risks and further reducing negative impacts.

Description

Technical Field

[0001] The present application relates to the field of computer application technology, and in particular to an information dissemination risk prediction method and device, a storage medium and an electronic device. Background Art

[0002] In recent years, with the development and changes in communication technology, means and environment, the channels for publishing manuscripts have become more diversified, including newspapers, magazines, websites, social platforms, etc. The audience groups of the published manuscripts are huge and the coverage is wide. Once excessive public discussion is triggered during the process of information dissemination, it will bring great negative impact.

[0003] Therefore, how to provide a technical solution that can predict the information dissemination risks after the manuscript is published in order to gain the initiative to respond and thus reduce the negative impact is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0004] The present application provides a method and device for predicting information dissemination risks, a storage medium and an electronic device, the purpose of which is to predict the information dissemination risks after the manuscript is published, so as to gain the initiative to respond and thus reduce negative impacts.

[0005] In order to achieve the above objectives, this application provides the following technical solutions:

[0006] A method for predicting information dissemination risk, comprising:

[0007] After the manuscript to be predicted is published, the comment data generated by the manuscript to be predicted in each publishing channel is obtained according to a preset period;

[0008] Input the comment data generated by each publishing channel into the pre-built prediction model to obtain the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point;

[0009] Based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, a trend graph of the manuscript to be predicted is drawn; the trend graph includes a trend graph of the total number of comments and a trend graph of the number of negative comments;

[0010] Based on the trend graph of the manuscript to be predicted, determining a target alarm threshold from various preset alarm thresholds;

[0011] Based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, calculate the proportion of negative comments for the manuscript to be predicted;

[0012] If the proportion of negative comments is not greater than the target warning threshold, it is determined that the manuscript to be predicted does not have information dissemination risk;

[0013] If the proportion of negative comments is greater than the target warning threshold, it is determined that the predicted manuscript has information dissemination risks and an alarm message is issued.

[0014] In the above method, optionally, drawing a trend graph of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point includes:

[0015] Based on the predicted value of the total number of comments for each publishing channel at each future time point, calculate the total predicted value of the total number of comments at each future time point;

[0016] Based on the predicted number of negative comments for each publishing channel at each future time point, calculate the total predicted number of negative comments at each future time point;

[0017] Based on the predicted total value of the total number of comments at each future time point, draw a trend graph of the total number of comments for the manuscript to be predicted;

[0018] Based on the predicted total value of the number of negative comments at each future time point, a trend graph of the number of negative comments for the manuscript to be predicted is drawn.

[0019] Optionally, the method described above, based on the trend graph of the manuscript to be predicted, determining the target alarm threshold from various preset alarm thresholds includes:

[0020] Based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph and each historical trend graph to obtain a calculation result;

[0021] Based on the calculation result, determining the classification result to which the manuscript to be predicted belongs in the pre-constructed clustering tree;

[0022] The alarm threshold corresponding to the classification result among the preset alarm thresholds is determined as the target alarm threshold.

[0023] In the above method, optionally, the process of constructing the prediction model includes:

[0024] Acquire a training data set and a test data set; the training data set includes a training data sequence of the historical manuscript in each publishing channel, the training data sequence includes a training data combination corresponding to multiple time points, the training data combination includes a time point, a publishing channel, the total number of historical comments and the number of negative comments of the historical manuscript in the publishing channel and at the time point, and the test data set includes a test data sequence of the historical manuscript in each publishing channel;

[0025] Start the preset counter to perform initial counting;

[0026] Perform a stationarity check on each training data sequence;

[0027] If any training data sequence fails the stationarity check, performing differential processing on each training data sequence, adding one to the counting result of the counter, and returning to perform the step of performing stationarity check on each training data sequence based on each training data sequence after the differential processing;

[0028] If all training data sequences pass the stationarity check, the number of differences is determined based on the current counting result of the counter;

[0029] Based on each training data sequence, draw partial autocorrelation graphs and autocorrelation graphs;

[0030] Analyze the tailing of the autocorrelation graph and the partial autocorrelation graph to obtain the value range of the number of autoregressive items and the value range of the number of moving average items;

[0031] Combining the values ​​included in the value range of the autoregressive term number and the values ​​included in the value range of the moving average term number in pairs, to obtain each combination result; each combination result includes the number of autoregressive terms and the number of moving average terms;

[0032] For each combination result, based on the combination result and the number of differences, a differential autoregressive moving average model corresponding to the combination result is constructed;

[0033] Using the AIC criterion, the target difference autoregressive moving average model is determined from each difference autoregressive moving average model;

[0034] Based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result;

[0035] The error of the test result is calculated, and if the error is less than a preset threshold, the target differential autoregressive average model is used as a prediction model.

[0036] In the above method, optionally, the step of performing a stationarity check on each training data sequence includes:

[0037] For each training data sequence, determine whether there is a unit root in the training data sequence. If there is a unit root in the training data sequence, determine that the training data sequence fails the stationarity check. If there is no unit root in the training data sequence, determine that the training data sequence passes the stationarity check.

[0038] The above method, optionally, after determining the target differential autoregressive moving average model from each differential autoregressive moving average model using the AIC criterion, further comprises:

[0039] The target difference autoregressive moving average model is tested.

[0040] In the above method, optionally, the calculation of the proportion of negative comments of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point includes:

[0041] Get the weight of each publishing channel;

[0042] Based on the weight of each publishing channel, the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the proportion of negative comments for the manuscript to be predicted is calculated.

[0043] An information dissemination risk prediction device, comprising:

[0044] An acquisition unit is used to acquire the comment data generated by the manuscript to be predicted in each publishing channel according to a preset period after the manuscript to be predicted is published;

[0045] An input unit, used to input the comment data generated by each publishing channel into a pre-built prediction model, to obtain a predicted value of the total number of comments and a predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point;

[0046] A drawing unit, used to draw a trend graph of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point; the trend graph includes a trend graph of the total number of comments and a trend graph of the number of negative comments;

[0047] A first determining unit, configured to determine a target alarm threshold from various preset alarm thresholds based on the trend graph of the manuscript to be predicted;

[0048] A calculation unit, used to calculate the proportion of negative comments of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point;

[0049] A second determination unit is configured to determine that the manuscript to be predicted does not have information dissemination risk if the proportion of negative comments is not greater than the target warning threshold;

[0050] The third determination unit is used to determine that the predicted manuscript has information dissemination risks and issue a warning message if the proportion of the negative comments is greater than the target warning threshold.

[0051] A storage medium stores an instruction set, wherein the instruction set, when executed by a processor, implements the information propagation risk prediction method as described above.

[0052] An electronic device, comprising:

[0053] A memory for storing at least one set of instructions;

[0054] A processor is used to execute an instruction set stored in the memory, and implement the above-mentioned information propagation risk prediction method by executing the instruction set.

[0055] Compared with the prior art, this application has the following advantages:

[0056] The present application provides a method and device for predicting information dissemination risk, a storage medium and an electronic device, the method comprising: obtaining the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point based on the comment data generated by the manuscript to be predicted in each publishing channel according to a preset cycle, drawing a trend chart of the total number of comments and a trend chart of the number of negative comments, and calculating the proportion of negative comments, and determining the target alarm threshold based on the trend chart of the total number of comments and the trend chart of the number of negative comments. If the proportion of negative comments is not greater than the target alarm threshold, it is determined that the manuscript to be predicted does not have an information dissemination risk, otherwise, it is determined that the manuscript to be predicted has an information dissemination risk. It can be seen that the present application scheme realizes the prediction of information dissemination risk, provides more response time for handling risks, thereby reducing information dissemination risks and further reducing negative impacts. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0058] Figure 1 A method flow chart of an information dissemination risk prediction method provided in this application;

[0059] Figure 2 Another method flow chart of an information dissemination risk prediction method provided by this application;

[0060] Figure 3 An example diagram of an information dissemination risk prediction method provided in this application;

[0061] Figure 4 A flowchart of another method for predicting information dissemination risks provided by the present application;

[0062] Figure 5 Another method flow chart of an information dissemination risk prediction method provided by this application;

[0063] Figure 6 A schematic diagram of the structure of an information dissemination risk prediction device provided in this application;

[0064] Figure 7 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0066] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0067] It should be noted that the concepts such as "first" and "second" mentioned in the present application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0068] It should be noted that the modifications of "one" and "plurality" mentioned in the disclosure of this application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0069] The present application can be used in many general or special computing device environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multi-processor devices, distributed computing environments including any of the above devices or equipment, etc.

[0070] The present application embodiment provides a method for predicting information propagation risk. The method can be applied to a variety of system platforms, and its execution subject can be a processor running on various mobile devices. The flowchart of the method is as follows: Figure 1 As shown, specifically including:

[0071] S101. After the manuscript to be predicted is published, the comment data generated by the manuscript to be predicted in each publishing channel is obtained according to a preset period.

[0072] In this embodiment, after the manuscript to be predicted is released, the comment data generated by the manuscript to be predicted in each release channel is obtained according to a preset period, wherein the comment data at least includes the total number of comments and the number of negative comments. That is, after the manuscript to be predicted is released, the comment data generated by the manuscript to be predicted in each release channel is obtained at each preset time interval.

[0073] Exemplary publishing channels include, but are not limited to, newspapers, magazines, websites, and social platforms.

[0074] S102, inputting the comment data generated by each publishing channel into a pre-built prediction model to obtain the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point.

[0075] In this embodiment, a prediction model is constructed in advance.

[0076] See also Figure 2 ,The process of building the prediction model includes the following steps:

[0077] S201. Obtain a training data set and a test data set.

[0078] In this embodiment, a training data set and a test data set are obtained, wherein the training data set includes a training data sequence of historical manuscripts in each publishing channel, the training data sequence includes a training data combination of multiple time points, and the training data combination includes a time point, a publishing channel, and the historical total comment data and the number of negative comments of the historical manuscripts in the publishing channel and time point. That is to say, each training data combination includes data of four dimensions, specifically including a time point, a publishing channel, the historical total comment data and the number of negative comments of the historical manuscripts in the publishing channel and time point.

[0079] Among them, each test data set includes a test data sequence of historical manuscripts in each publishing channel, the test data sequence includes a training data combination at multiple time points, and the training data combination includes the time point, publishing channel, historical total comment data and the number of negative comments of historical manuscripts in the publishing channel and time point.

[0080] S202: Start a preset counter to perform initial counting.

[0081] In this example, a preset counter is started to count to one.

[0082] S203: Perform a stationarity check on each training data sequence.

[0083] In this embodiment, a stationarity check is performed on each training data sequence. Specifically, a unit root test method can be used to perform a stationarity check on each training data sequence.

[0084] In this embodiment, the process of performing stationarity check on each training data sequence specifically includes the following steps:

[0085] For each training data sequence, determine whether there is a unit root in the training data sequence. If there is a unit root in the training data sequence, it is determined that the training data sequence fails the stationarity check. If there is no unit root in the training data sequence, it is determined that the training data sequence passes the stationarity check.

[0086] S204: Determine whether all training data sequences have passed the stationarity check. If so, execute S205; if not, execute S206.

[0087] S205: Determine the number of differential times based on the current counting result of the counter.

[0088] In this embodiment, if all training data sequences pass the stationarity check, the number of differences is determined based on the current counting result of the timer, that is, the parameter d of the ARIMA (Autoregressive Integrated Moving Average model) model is determined.

[0089] It should be noted that the current counting result of the counter is the same as the order of the final differential processing. Therefore, the determined parameter d is equal to the order of the final differential processing.

[0090] In this embodiment, if all training data sequences pass the stationarity check when the stationarity check is performed for the first time, the current count result of the counter is 1, and the parameter d is also 1.

[0091] S206 , performing differential processing on each training data sequence, adding one to the count of the counter, and returning to execute S203 based on each training data sequence after the differential processing.

[0092] In this embodiment, if any training data sequence fails to pass the differential processing, differential processing is performed on each training data sequence. Specifically, N-order differential processing is performed on each training data sequence, where N is a positive integer, and N is determined based on the current counting result of the counter. Specifically, N is equal to the current counting result of the counter plus one.

[0093] In this embodiment, after differential processing is performed on each training data sequence, the counting result of the counter is incremented by one, and based on each training data sequence after differential processing, the process returns to execute step S203.

[0094] S207. Based on each training data sequence, draw a partial autocorrelation graph and an autocorrelation graph.

[0095] In this embodiment, after the number of differences is determined, partial autocorrelation graphs and autocorrelation graphs are drawn based on each training data sequence.

[0096] For example, the autocorrelation diagram obtained is plotted as Figure 3 As shown in Figure a, the partial autocorrelation diagram obtained is as follows Figure 3 As shown in Figure b.

[0097] S208, analyzing the tailing of the autocorrelation graph and the partial autocorrelation graph to obtain the value range of the number of autoregressive items and the value range of the number of moving average items.

[0098] In this embodiment, the tailing of the autocorrelation graph is analyzed to obtain the value range of the number of autoregressive terms, that is, to obtain the parameter p of the ARIMA model; the tailing of the partial autocorrelation graph is analyzed to obtain the value range of the number of moving average terms, that is, to obtain the parameter q of the ARIMA model.

[0099] Specifically, determine the function values ​​at different lag orders lab in the autocorrelation diagram and the partial autocorrelation diagram, determine the lab corresponding to the function value that is not in the preset interval from each function value, and based on the lab corresponding to the function value that is not in the preset interval, determine the value range of the number of autoregressive terms and the value range of the number of moving average terms. Preferably, the preset interval is [-0.2, 0.2].

[0100] See also Figure 3 , the tailing analysis process of partial autocorrelation graph and autocorrelation graph is illustrated as follows:

[0101] Figure 3 Figure a is the autocorrelation graph, and figure b is the partial autocorrelation graph.

[0102] In Figure a, at lag=1, 8, and 9, it is significantly not 0, that is, it is not within the preset interval [-0.2, 0.2]. It can be determined that p can be 1, 2, or 3.

[0103] In Figure b, at lag = 1, 2, 3, 4, 5, 7, it is significantly not 0, that is, it is not in the preset interval [-0.2, 0.2]. It can be determined that q can be 1, 2, 3, 4, 5, 6.

[0104] S209, combining the values ​​included in the value range of the number of autoregressive items and the values ​​included in the value range of the number of moving average items in pairs, to obtain each combination result.

[0105] In this embodiment, the values ​​included in the value range of the number of autoregressive terms and the values ​​included in the value range of the number of moving average terms are combined in pairs to obtain each combination result; each combination result includes the number of autoregressive terms and the number of moving average terms.

[0106] For example, the range of values ​​for the number of autoregressive terms includes the values ​​1 and 2, and the range of values ​​for the number of moving average terms includes the values ​​1, 2, and 3. Then the results of pairwise combinations are (1,1), (1,2), (1,3), (2,1), (1,2), and (2,3).

[0107] S210: For each combination result, based on the combination result and the number of differences, construct a differential autoregressive moving average model corresponding to the combination result.

[0108] In this embodiment, for each combination result, a differential autoregressive moving average model corresponding to the combination result is constructed based on the combination result and the number of differences, that is, based on the number of differences, the number of autoregressive items and the number of moving average items included in the combination result, an ARIMA model corresponding to the combination result is constructed.

[0109] S211. Determine a target differential autoregressive moving average model from each differential autoregressive moving average model using the AIC criterion.

[0110] In this embodiment, the target differential autoregressive moving average model is determined from various differential autoregressive moving average models using the AIC (Akaike information criterion) criterion.

[0111] Specifically, for each differential autoregressive moving average model, based on the training data sequence, the AIC value of the differential autoregressive moving average model is calculated by a preset formula, wherein the preset formula is:

[0112] AIC=-2log(L)+2K

[0113] Among them, AIC is the AIC value of the difference autoregressive moving average model, L is the maximum likelihood estimate of the difference autoregressive moving average model, and K is the dimension of the training data sequence.

[0114] In this embodiment, based on the AIC values ​​of the differential autoregressive moving average models, the differential autoregressive moving average model corresponding to the minimum AIC value is determined, and the differential autoregressive moving average model corresponding to the minimum AIC value is determined as the target differential autoregressive moving average model.

[0115] In this embodiment, optionally, after determining the target differential autoregressive moving average model from each differential autoregressive moving average model, the method further includes:

[0116] The target difference autoregressive moving average model is tested.

[0117] In this embodiment, the target differential autoregressive moving average model is tested. Exemplarily, the target differential autoregressive moving average model may be tested for autocorrelation, normality or white noise.

[0118] In this embodiment, a residual time series diagram of the target differential autoregressive moving average model is drawn. Based on the residual time series diagram, within the 20th order, the p-value is significantly greater than 0.05, which means that the residuals of the target differential autoregressive moving average model have good independence. The distribution of the residual points of the target differential autoregressive moving average model is close to the index straight line except for the errors at both ends, indicating that the residuals basically conform to the normal distribution.

[0119] S212: Based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result.

[0120] In this embodiment, based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result. Specifically, each test data sequence is predicted using the target differential autoregressive moving average model, and the prediction result is the test result.

[0121] S213, calculating the error of the test result, if the error is less than a preset threshold, using the target differential autoregressive average model as a prediction model.

[0122] In this embodiment, the error of the test result is calculated. Specifically, based on the test result and the test data sequence, the root mean square error and the mean absolute percentage error are calculated.

[0123] In this embodiment, it is determined whether the error is less than a preset threshold. Specifically, it is determined whether the root mean square error is less than a preset first error threshold, and whether the mean absolute percentage error is less than a preset second error threshold. If the root mean square error is less than the preset first error threshold, and the mean absolute percentage error is less than the preset second error threshold, the target differential autoregressive average is used as the prediction model.

[0124] The information dissemination risk prediction method provided in the embodiment of the present application has data in the training data sequence coming from different publishing channels, with relatively comprehensive coverage, and the data is authentic and reliable.

[0125] In this embodiment, the comment data generated by each publishing channel is input into a pre-built prediction model. After the prediction model processes the comment data, the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point are obtained.

[0126] S103. Draw a trend chart of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point.

[0127] In this embodiment, based on the predicted total average number and the predicted number of negative comments of each publishing channel at each future time point, a trend chart of the manuscript to be predicted is drawn, wherein the trend chart includes a trend chart of the total number of comments and a trend chart of the negative average number.

[0128] In this embodiment, since the generated comment data is obtained according to a preset period, the trend chart of the manuscript to be predicted will be adjusted according to the generated comment data, and has a strong timeliness.

[0129] See also Figure 4 , based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the process of drawing a trend chart of the manuscript to be predicted specifically includes the following steps:

[0130] S401. Calculate the predicted total number of comments at each future time point based on the predicted total number of comments for each publishing channel at each future time point.

[0131] In this embodiment, based on the predicted value of the total number of comments for each publishing channel at each future time point, the predicted total number of comments at each future time point is calculated. Specifically, for each future time point, the predicted value of the total number of comments for each publishing channel at that future time point is accumulated to obtain the predicted total number of comments at that future time point.

[0132] S402. Calculate the predicted total value of the number of negative comments at each future time point based on the predicted value of the number of negative comments at each publishing channel at each future time point.

[0133] In this embodiment, based on the predicted value of the number of negative comments for each publishing channel at each future time point, the predicted total value of negative comment data at each future time point is calculated. Specifically, for each future time point, the predicted average negative value of each publishing channel at that future time point is accumulated to obtain the predicted total value of the number of negative comments at that future time point.

[0134] S403. Based on the predicted total value of the total number of comments at each future time point, draw a trend chart of the total number of comments for the manuscript to be predicted.

[0135] S404. Based on the predicted total value of the number of negative comments at each future time point, draw a trend graph of the number of negative comments of the manuscript to be predicted.

[0136] S104: Based on the trend graph of the manuscript to be predicted, determine a target alarm threshold from various preset alarm thresholds.

[0137] In this embodiment, a clustering tree is pre-constructed, and the clustering tree includes a plurality of classification results, and each classification result corresponds to a preset alarm threshold.

[0138] The process of building a clustering tree includes:

[0139] The dynamic time warping distance (DTW) is used to calculate the pairwise similarity distances of each training data sequence corresponding to the historical manuscripts. Based on the results of the similarity distance calculation, each preset hierarchical clustering algorithm is used to perform hierarchical clustering on each training data sequence to obtain the initial clustering tree corresponding to each hierarchical clustering algorithm, and the Shepart diagram of each initial clustering tree is drawn. Based on the Shepart diagram of each initial clustering tree, a clustering tree is determined from each initial clustering tree.

[0140] In this embodiment, if the fit between the same phenotype distance and the original object distance is more consistent, it indicates that the clustering model is more suitable.

[0141] Exemplarily, the hierarchical clustering algorithms include average distance method (average), centroid method (centroid), median distance method (median), shortest distance method (single), longest distance method (complete), minimum variance method (ward.D), and arithmetic mean weight method (mcquitty).

[0142] In this embodiment, an evaluation method may also be used to evaluate the clustering result of the clustering tree. The evaluation method includes an intrinsic method and an extrinsic method.

[0143] Intrinsic method: The silhouette coefficient is used for evaluation. The value of the silhouette coefficient is between -1 and 1. When the silhouette coefficient is close to 1, the cluster containing a sequence is compact and the sequence is far away from other clusters. When the value of the silhouette coefficient is negative, the sequence is closer to objects in other clusters than to objects in the same cluster as itself.

[0144] External method: The clustering results are compared with the manual classification to evaluate. The evaluation indicators used are JC, FMI, and RI. Assume that the data set D = {x 1 ,x 2 ,…,xn}, assuming that the clusters obtained by clustering are divided into labels p = [p 1 ,…,p n ], the real label is r = [r 1 ,…,r n ], consider the samples in pairs, and we have the following formula:

[0145]

[0146] Among them, SS contains sample pairs predicted to be the same cluster and with the same true labels, SD contains sample pairs predicted to be the same cluster but with inconsistent true labels, DS contains sample pairs predicted to be different clusters but with consistent true labels, and DD contains sample pairs predicted to be different clusters and with inconsistent true labels.

[0147] Based on the above formulas (1) to (4), the indicators are as shown below. The result values ​​of these indicators are all in the interval [0,1], and the larger the value, the better. The clustering method with better overall performance of the three indicators is selected as the standard for cluster distance calculation.

[0148]

[0149] In this embodiment, based on the trend graph of the manuscript to be predicted, a target alarm threshold is determined from various preset alarm thresholds.

[0150] See also Figure 5 , based on the trend graph of the manuscript to be predicted, the process of determining the target alarm threshold from various preset alarm thresholds specifically includes the following steps:

[0151] S501. Based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph and each historical trend graph to obtain a calculation result.

[0152] In this embodiment, a plurality of historical trend graphs are preset. Specifically, the historical trend graphs are constructed based on training data sequences of historical manuscripts.

[0153] In this embodiment, based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph of the manuscript to be predicted and each historical trend graph to obtain a calculation result.

[0154] Exemplarily, the distance calculation algorithm may be a dynamic time warping distance DTW algorithm.

[0155] S502: Based on the calculation result, determine the classification result to which the manuscript to be predicted belongs in the pre-constructed clustering tree.

[0156] In this embodiment, based on the calculation result, the classification result to which the manuscript to be predicted belongs in the pre-constructed clustering tree is determined.

[0157] S503: Determine the alarm threshold corresponding to the classification result among the preset alarm thresholds as the target alarm threshold.

[0158] In this embodiment, the alarm threshold corresponding to the classification result among the preset alarm thresholds is determined as the target alarm threshold.

[0159] S105. Calculate the proportion of negative comments in the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point.

[0160] In this embodiment, based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the proportion of negative comments of the manuscript to be predicted is calculated.

[0161] Specifically, the following steps are included:

[0162] Get the weight of each publishing channel;

[0163] Based on the weight of each publishing channel, the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the proportion of negative comments for the manuscript to be predicted is calculated.

[0164] In this embodiment, a weight of each publishing channel is preset, and the preset weight of each publishing channel is obtained.

[0165] In this embodiment, the weight of each publishing channel is related to the media influence and the number of fans following the publishing channel, wherein the greater the media influence and the more fans following the channel, the greater the weight.

[0166] In this embodiment, based on the weight of each publishing channel, the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point, the proportion of negative comments of the manuscript to be predicted is calculated.

[0167] S106, determine whether the proportion of negative comments is greater than the target alarm threshold, if so, execute S107, if not, execute S108.

[0168] S107. Determine that the predicted manuscript has information dissemination risks and issue a warning message.

[0169] In this embodiment, if the proportion of negative comments is greater than the target warning threshold, it is determined that the manuscript to be predicted has information dissemination risks, and an alarm message is issued.

[0170] In this embodiment, the warning information may be issued in the form of outputting prompt text, or in the form of outputting prompt sound.

[0171] S108. Determine that there is no risk of information dissemination for the manuscript to be predicted.

[0172] In this embodiment, if the proportion of negative comments is not greater than the target alarm threshold, it is determined that there is no information dissemination risk for the alarm to be predicted.

[0173] The information dissemination risk prediction method provided by the embodiment of the present application, after the manuscript to be predicted is published, according to the preset cycle, based on the comment data generated by the manuscript to be predicted in each publishing channel, through the pre-built prediction model, the total number of comments and the predicted number of negative comments corresponding to each publishing channel at each future time point are obtained, and based on the total number of comments and the predicted number of negative comments of each publishing channel at each future time point, the trend graph of the total number of comments and the trend graph of the negative number of comments are drawn, and the proportion of negative comments is calculated. Based on the trend graph of the total number of comments and the trend graph of the negative number of comments, the target alarm threshold is determined from the preset alarm thresholds. If the proportion of negative comments is not greater than the target alarm threshold, it is determined that the manuscript to be predicted does not have information dissemination risk. If the proportion of negative comments is greater than the target alarm threshold, it is determined that the manuscript to be predicted has information dissemination risk. It can be seen that the present application scheme realizes the prediction of information dissemination risk, provides more response time for handling risks, thereby reducing information dissemination risks, and then reducing negative impacts. In addition, the trend graph of the manuscript to be predicted is adjusted with the number of comments that have been generated, and has strong timeliness.

[0174] It should be noted that although the operations are depicted in a particular order, this should not be understood as requiring that the operations be performed in the particular order shown or in a sequential order. In certain circumstances, multitasking and parallel processing may be advantageous.

[0175] It should be understood that the various steps described in the method embodiments disclosed in the present application can be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0176] and Figure 1 Corresponding to the method described above, the embodiment of the present application also provides an information propagation risk prediction device for Figure 1 The specific implementation of the method is shown in the following diagram: Figure 6 As shown, specifically including:

[0177] The acquisition unit 601 is used to acquire the comment data generated by the manuscript to be predicted in each publishing channel according to a preset period after the manuscript to be predicted is published;

[0178] An input unit 602 is used to input the comment data generated by each publishing channel into a pre-built prediction model to obtain a predicted value of the total number of comments and a predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point;

[0179] A drawing unit 603 is used to draw a trend graph of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point; the trend graph includes a trend graph of the total number of comments and a trend graph of the number of negative comments;

[0180] A first determining unit 604 is configured to determine a target alarm threshold from various preset alarm thresholds based on the trend graph of the manuscript to be predicted;

[0181] A calculation unit 605 is used to calculate the proportion of negative comments of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point;

[0182] A second determination unit 606 is configured to determine that the manuscript to be predicted does not have information dissemination risk if the proportion of negative comments is not greater than the target warning threshold;

[0183] The third determination unit 607 is used to determine that the predicted manuscript has information dissemination risks and issue a warning message if the proportion of the negative comments is greater than the target warning threshold.

[0184] The information dissemination risk prediction device provided in the embodiment of the present application realizes the prediction of information dissemination risks, provides more response time for handling risks, thereby reducing information dissemination risks and further reducing negative impacts. In addition, the trend graph of the manuscript to be predicted is adjusted according to the number of comments generated, and has strong timeliness.

[0185] In one embodiment of the present application, based on the above solution, the drawing unit 603 is specifically used for:

[0186] Based on the predicted value of the total number of comments for each publishing channel at each future time point, calculate the total predicted value of the total number of comments at each future time point;

[0187] Based on the predicted number of negative comments for each publishing channel at each future time point, calculate the total predicted number of negative comments at each future time point;

[0188] Based on the predicted total value of the total number of comments at each future time point, draw a trend graph of the total number of comments for the manuscript to be predicted;

[0189] Based on the predicted total value of the number of negative comments at each future time point, a trend graph of the number of negative comments for the manuscript to be predicted is drawn.

[0190] In one embodiment of the present application, based on the above solution, the first determining unit 604 is specifically configured to:

[0191] Based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph and each historical trend graph to obtain a calculation result;

[0192] Based on the calculation result, determining the classification result to which the manuscript to be predicted belongs in the pre-constructed clustering tree;

[0193] The alarm threshold corresponding to the classification result among the preset alarm thresholds is determined as the target alarm threshold.

[0194] In one embodiment of the present application, based on the above solution, the input unit 602 is specifically used in the process of building the prediction model to:

[0195] Acquire a training data set and a test data set; the training data set includes a training data sequence of the historical manuscript in each publishing channel, the training data sequence includes a training data combination corresponding to multiple time points, the training data combination includes a time point, a publishing channel, the total number of historical comments and the number of negative comments of the historical manuscript in the publishing channel and at the time point, and the test data set includes a test data sequence of the historical manuscript in each publishing channel;

[0196] Start the preset counter to perform initial counting;

[0197] Perform a stationarity check on each training data sequence;

[0198] If any training data sequence fails the stationarity check, performing differential processing on each training data sequence, adding one to the counting result of the counter, and returning to perform the step of performing stationarity check on each training data sequence based on each training data sequence after the differential processing;

[0199] If all training data sequences pass the stationarity check, the number of differences is determined based on the current counting result of the counter;

[0200] Based on each training data sequence, draw partial autocorrelation graphs and autocorrelation graphs;

[0201] Analyze the tailing of the autocorrelation graph and the partial autocorrelation graph to obtain the value range of the number of autoregressive items and the value range of the number of moving average items;

[0202] Combining the values ​​included in the value range of the autoregressive term number and the values ​​included in the value range of the moving average term number in pairs, to obtain each combination result; each combination result includes the number of autoregressive terms and the number of moving average terms;

[0203] For each combination result, based on the combination result and the number of differences, a differential autoregressive moving average model corresponding to the combination result is constructed;

[0204] Using the AIC criterion, the target difference autoregressive moving average model is determined from each difference autoregressive moving average model;

[0205] Based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result;

[0206] The error of the test result is calculated, and if the error is less than a preset threshold, the target differential autoregressive average model is used as a prediction model.

[0207] In one embodiment of the present application, based on the above solution, when the input unit 604 performs a stationarity check on each training data sequence, it is specifically used to:

[0208] For each training data sequence, determine whether there is a unit root in the training data sequence. If there is a unit root in the training data sequence, determine that the training data sequence fails the stationarity check. If there is no unit root in the training data sequence, determine that the training data sequence passes the stationarity check.

[0209] In one embodiment of the present application, based on the above solution, the input unit 604 is further used to:

[0210] The target difference autoregressive moving average model is tested.

[0211] In one embodiment of the present application, based on the above solution, the calculation unit 605 is specifically used to:

[0212] Get the weight of each publishing channel;

[0213] Based on the weight of each publishing channel, the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the proportion of negative comments for the manuscript to be predicted is calculated.

[0214] An embodiment of the present application further provides a storage medium storing an instruction set, wherein when the instruction set is executed, the information propagation risk prediction method disclosed in any of the above embodiments is executed.

[0215] The present application also provides an electronic device, whose structural diagram is as follows: Figure 7 As shown, it specifically includes a memory 701 for storing at least one set of instruction sets; a processor 702 for executing the instruction set stored in the memory, and implementing the information propagation risk prediction method disclosed in any of the above embodiments by executing the instruction set.

[0216] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

[0217] Although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope disclosed in the present application. Certain features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.

[0218] The above description is only a preferred embodiment disclosed in this application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the disclosure of this application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the disclosure of this application (but not limited to) to form a technical solution.

Claims

1. A method for predicting information propagation risk, characterized in that: include: After the manuscript to be predicted is published, the comment data generated by the manuscript to be predicted in each publishing channel is obtained according to a preset period; Input the comment data generated by each publishing channel into the pre-built prediction model to obtain the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point; Based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, draw a trend graph of the manuscript to be predicted; The trend graph includes a total number of comments trend graph and a negative number of comments trend graph; Based on the trend graph of the manuscript to be predicted, determining a target alarm threshold from various preset alarm thresholds; Based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, calculate the proportion of negative comments for the manuscript to be predicted; If the proportion of negative comments is not greater than the target warning threshold, it is determined that the manuscript to be predicted does not have information dissemination risk; If the proportion of negative comments is greater than the target warning threshold, it is determined that the predicted manuscript has information dissemination risks and a warning message is issued; Wherein, determining a target alarm threshold from various preset alarm thresholds based on the trend graph of the manuscript to be predicted includes: Based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph and each historical trend graph to obtain a calculation result; Based on the calculation result, determining the classification result to which the manuscript to be predicted belongs in a pre-constructed clustering tree; the clustering tree includes a plurality of classification results, each of which corresponds to a preset alarm threshold; Determining the alarm threshold corresponding to the classification result among the preset alarm thresholds as the target alarm threshold; The process of constructing the prediction model includes: Acquire a training data set and a test data set; the training data set includes a training data sequence of the historical manuscript in each publishing channel, the training data sequence includes a training data combination corresponding to multiple time points, the training data combination includes a time point, a publishing channel, the total number of historical comments and the number of negative comments of the historical manuscript in the publishing channel and at the time point, and the test data set includes a test data sequence of the historical manuscript in each publishing channel; Start the preset counter to perform initial counting; Perform a stationarity check on each training data sequence; If any training data sequence fails the stationarity check, performing differential processing on each training data sequence, adding one to the counting result of the counter, and returning to perform the step of performing stationarity check on each training data sequence based on each training data sequence after the differential processing; If all training data sequences pass the stationarity check, the number of differences is determined based on the current counting result of the counter; Based on each training data sequence, draw partial autocorrelation graphs and autocorrelation graphs; Analyze the tailing of the autocorrelation graph and the partial autocorrelation graph to obtain the value range of the number of autoregressive items and the value range of the number of moving average items; Combining the values ​​included in the value range of the autoregressive term number and the values ​​included in the value range of the moving average term number in pairs, to obtain each combination result; each combination result includes the number of autoregressive terms and the number of moving average terms; For each combination result, based on the combination result and the number of differences, a differential autoregressive moving average model corresponding to the combination result is constructed; Using the AIC criterion, the target difference autoregressive moving average model is determined from each difference autoregressive moving average model; Based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result; The error of the test result is calculated, and if the error is less than a preset threshold, the target differential autoregressive moving average model is used as a prediction model.

2. The method according to claim 1, characterized in that The method of drawing a trend graph of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point includes: Based on the predicted value of the total number of comments for each publishing channel at each future time point, calculate the total predicted value of the total number of comments at each future time point; Based on the predicted number of negative comments for each publishing channel at each future time point, calculate the total predicted number of negative comments at each future time point; Based on the predicted total value of the total number of comments at each future time point, draw a trend graph of the total number of comments for the manuscript to be predicted; Based on the predicted total value of the number of negative comments at each future time point, a trend graph of the number of negative comments of the manuscript to be predicted is drawn.

3. The method according to claim 1, characterized in that The stationarity check of each training data sequence includes: For each training data sequence, determine whether there is a unit root in the training data sequence. If there is a unit root in the training data sequence, determine that the training data sequence fails the stationarity check. If there is no unit root in the training data sequence, determine that the training data sequence passes the stationarity check.

4. The method according to claim 1, characterized in that: After determining the target differential autoregressive moving average model from each differential autoregressive moving average model by using the AIC criterion, the method further includes: The target difference autoregressive moving average model is tested.

5. The method according to claim 1, characterized in that The calculation of the proportion of negative comments of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point includes: Get the weight of each publishing channel; Based on the weight of each publishing channel, the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point, the proportion of negative comments for the manuscript to be predicted is calculated.

6. An information dissemination risk prediction device, characterized in that: include: An acquisition unit is used to acquire the comment data generated by the manuscript to be predicted in each publishing channel according to a preset period after the manuscript to be predicted is published; An input unit, used to input the comment data generated by each publishing channel into a pre-built prediction model, to obtain a predicted value of the total number of comments and a predicted value of the number of negative comments for each publishing channel corresponding to the manuscript to be predicted at each future time point; A drawing unit, used for drawing a trend graph of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments of each publishing channel at each future time point; The trend graph includes a total number of comments trend graph and a negative number of comments trend graph; A first determining unit, configured to determine a target alarm threshold from various preset alarm thresholds based on the trend graph of the manuscript to be predicted; A calculation unit, used to calculate the proportion of negative comments of the manuscript to be predicted based on the predicted value of the total number of comments and the predicted value of the number of negative comments for each publishing channel at each future time point; A second determination unit is configured to determine that the manuscript to be predicted does not have information dissemination risk if the proportion of negative comments is not greater than the target warning threshold; A third determination unit is configured to determine that the predicted manuscript has information dissemination risks and issue a warning message if the proportion of the negative comments is greater than a target warning threshold; Wherein, the first determining unit is specifically used to: Based on the trend graph of the manuscript to be predicted and each preset historical trend graph, a preset distance calculation algorithm is used to calculate the distance between the trend graph and each historical trend graph to obtain a calculation result; Based on the calculation result, determining the classification result to which the manuscript to be predicted belongs in a pre-constructed clustering tree; the clustering tree includes a plurality of classification results, each of which corresponds to a preset alarm threshold; Determining the alarm threshold corresponding to the classification result among the preset alarm thresholds as the target alarm threshold; The process of constructing the prediction model includes: Acquire a training data set and a test data set; the training data set includes a training data sequence of the historical manuscript in each publishing channel, the training data sequence includes a training data combination corresponding to multiple time points, the training data combination includes a time point, a publishing channel, the total number of historical comments and the number of negative comments of the historical manuscript in the publishing channel and at the time point, and the test data set includes a test data sequence of the historical manuscript in each publishing channel; Start the preset counter to perform initial counting; Perform a stationarity check on each training data sequence; If any training data sequence fails the stationarity check, performing differential processing on each training data sequence, adding one to the counting result of the counter, and returning to perform the step of performing stationarity check on each training data sequence based on each training data sequence after the differential processing; If all training data sequences pass the stationarity check, the number of differences is determined based on the current counting result of the counter; Based on each training data sequence, draw partial autocorrelation graphs and autocorrelation graphs; Analyze the tailing of the autocorrelation graph and the partial autocorrelation graph to obtain the value range of the number of autoregressive items and the value range of the number of moving average items; Combining the values ​​included in the value range of the autoregressive term number and the values ​​included in the value range of the moving average term number in pairs, to obtain each combination result; each combination result includes the number of autoregressive terms and the number of moving average terms; For each combination result, based on the combination result and the number of differences, a differential autoregressive moving average model corresponding to the combination result is constructed; Using the AIC criterion, the target difference autoregressive moving average model is determined from each difference autoregressive moving average model; Based on each test data sequence, the target differential autoregressive moving average model is tested to obtain a test result; The error of the test result is calculated, and if the error is less than a preset threshold, the target differential autoregressive moving average model is used as a prediction model.

7. A storage medium, characterized in that: The storage medium stores an instruction set, wherein the instruction set, when executed by a processor, implements the information propagation risk prediction method according to any one of claims 1 to 5.

8. An electronic device, characterized in that: include: A memory for storing at least one set of instructions; A processor is used to execute an instruction set stored in the memory, and implement the information propagation risk prediction method as described in any one of claims 1 to 5 by executing the instruction set.

Images