A data anomaly alarm method and device

Abstract

The present invention discloses a data anomaly alarm method and device, relating to the field of computer technology, for resolving the problem of false alarms generated by existing monitoring methods. The method comprises: obtaining fluctuation data of a target currency pair within a first time period; determining a target parameter based on the fluctuation data of the target currency pair within the first time period and a preset neural network model; and generating an alarm message if the target parameter meets an alarm condition; the alarm condition is determined based on the neural network model and fluctuation data of the target currency pair within multiple second time periods, the multiple second time periods including a time period prior to the first time period. Embodiments of the present invention are applied to an alarm system for foreign exchange rates.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a data anomaly alarm method and device. Background Art

[0002] Currently, existing monitoring systems can monitor the updates and fluctuations of foreign exchange quotes. Specifically, after obtaining the update frequency or price volatility of a currency pair within a preset time period, the monitoring system compares it with a corresponding fixed threshold. If the update frequency of the currency pair falls below the fixed threshold, or the price volatility of the currency pair exceeds the fixed threshold, the monitoring system will issue an alarm.

[0003] However, due to the complexity of factors that cause changes in the foreign exchange market, existing monitoring methods often produce false alarms. Summary of the Invention

[0004] The embodiments of the present invention provide a data anomaly alarm method and device, which are used to solve the problem of false alarms generated by existing monitoring methods.

[0005] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0006] In a first aspect, a method for alarming for data anomalies is provided, comprising: obtaining fluctuation data of a target currency pair within a first time period; the fluctuation data within a time period includes a time identifier for reflecting whether a time period is within working hours, and the fluctuation data within a time period also includes the number of received updates and the number of sent updates of the target currency pair, and / or the received volatility and the sent volatility of the target currency pair; determining a target parameter based on the fluctuation data of the target currency pair within the first time period and a preset neural network model; the target parameter is used to reflect the number of updates and / or the volatility within a time period; if the target parameter meets an alarm condition, generating an alarm message; the alarm condition is determined based on the neural network model and the fluctuation data of the target currency pair in multiple second time periods, and the multiple second time periods include a time period before the first time period.

[0007] In a second aspect, an alarm device is provided, including an acquisition unit, a determination unit and a generation unit; the acquisition unit is used to acquire fluctuation data of a target currency pair within a first time period; the fluctuation data within a time period includes a time identifier for reflecting whether a time period is within working hours, and the fluctuation data within a time period also includes the number of received updates and the number of issued updates of the target currency pair, and / or the received volatility and the issued volatility of the target currency pair; the determination unit is used to determine a target parameter based on a preset neural network model and the fluctuation data of the target currency pair within the first time period acquired by the acquisition unit; the target parameter is used to reflect the number of updates and / or the volatility within a time period; the generation unit is used to generate an alarm message if the target parameter determined by the determination unit meets the alarm condition; the alarm condition is determined based on the neural network model and the fluctuation data of the target currency pair in multiple second time periods, and the multiple second time periods include the time period before the first time period.

[0008] In a third aspect, a computer-readable storage medium storing one or more programs is provided. The one or more programs include instructions that, when executed by a computer, cause the computer to execute the data anomaly alarm method of the first aspect.

[0009] In a fourth aspect, an alarm device comprises: a processor and a memory; wherein the memory is used to store one or more programs, and the one or more programs include computer execution instructions. When the alarm device is running, the processor executes the computer execution instructions stored in the memory to enable the alarm device to execute the data anomaly alarm method of the first aspect.

[0010] Embodiments of the present invention provide a data anomaly alarm method and device. First, an alarm device obtains fluctuation data for a target currency pair within a first time period. Based on the obtained fluctuation data and a preset neural network model, the device determines a target parameter reflecting the number of updates and / or volatility of the target currency pair within the time period. Because the fluctuation data within a time period includes a time stamp indicating whether the time period falls within working hours, and also includes the number of received and issued updates for the target currency pair, and / or the received and issued volatility of the target currency pair, compared to existing technologies, the device can utilize multi-dimensional information to determine the target parameter, which more comprehensively reflects the fluctuations of the target currency pair within the first time period. Furthermore, because the alarm device uses alarm conditions as the basis for determining whether to issue an alarm, and these alarm conditions are determined by the alarm device based on fluctuation data from multiple historical time periods under the same conditions, the determination criteria can be applied to the target parameters. Consequently, when determining whether to issue an alarm based on the target parameters and alarm conditions, the alarm result can be more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic structural diagram of an alarm system provided by an embodiment of the present invention;

[0012] Figure 2 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 1 ;

[0013] Figure 3 A schematic diagram of the loop structure of an LSTM neural network model provided by an embodiment of the present invention;

[0014] Figure 4 A schematic diagram of the structure of an LSTM unit provided in an embodiment of the present invention;

[0015] Figure 5 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 2 ;

[0016] Figure 6 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 3 ;

[0017] Figure 7 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 4 ;

[0018] Figure 8 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 5 ;

[0019] Figure 9 A data anomaly alarm method process diagram provided by an embodiment of the present invention Figure 6 ;

[0020] Figure 10 A schematic diagram of the structure of an alarm device provided in an embodiment of the present invention Figure 1 ;

[0021] Figure 11 A schematic diagram of the structure of an alarm device provided in an embodiment of the present invention Figure 2 ;

[0022] Figure 12 A schematic diagram of the structure of an alarm device provided in an embodiment of the present invention Figure 3 . DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0024] It should be noted that, in the embodiments of the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present invention should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0025] It should also be noted that, in the embodiments of the present invention, “of”, “corresponding”, “relevant” and “corresponding” can sometimes be used interchangeably. It should be pointed out that when the distinction between them is not emphasized, the meanings they intend to express are consistent.

[0026] In order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, words such as "first" and "second" are used to distinguish between identical or similar items with basically the same functions and effects. Those skilled in the art can understand that words such as "first" and "second" do not limit the quantity and execution order.

[0027] The data anomaly alarm method (hereinafter referred to as the alarm method) provided in the embodiment of the present invention can be applied to an alarm system. Figure 1 A structural diagram of the alarm system is shown in FIG. Figure 1 As shown, the alarm system 10 is used to monitor and alarm data anomalies. The alarm system 10 includes an alarm device 11 and a server 12. The alarm device 11 is connected to the server 12. The alarm device 11 and the server 12 can be connected in a wired manner or a wireless manner, which is not limited in the present embodiment.

[0028] The alarm device 11 can be used to interact with the server 12 for data, for example, to obtain the fluctuation data of the target currency pair in the first time period from the server 12 and send an alarm message to the server 12 .

[0029] The alarm device 11 may also be configured to determine a target parameter based on the fluctuation data of the target currency pair during a first time period and a preset neural network model, and to determine an alarm condition based on the fluctuation data of the target currency pair during multiple second time periods and the neural network model. Furthermore, the alarm device 11 may determine whether the target parameter satisfies the alarm condition based on the target parameter and alarm condition determined in the above steps, and further generate an alarm message based on the determination result.

[0030] In different application scenarios, the alarm device 11 and the server 12 may be independent devices or integrated into the same device, which is not specifically limited in the embodiment of the present invention.

[0031] When the alarm device 11 and the server 12 are integrated into the same device, the communication between the alarm device 11 and the server 12 is carried out as between the internal modules of the device. In this case, the communication process between the two is the same as the communication process between the alarm device 11 and the server 12 when they are independent of each other.

[0032] In the following embodiments provided by the embodiments of the present invention, the present disclosure is described by taking an example in which the alarm device 11 and the server 12 are independently configured.

[0033] Figure 2 The flowchart of a data anomaly alarm method according to some exemplary embodiments is shown. In some embodiments, the above data anomaly alarm method can be applied to Figure 1 The alarm device or server or other similar equipment shown.

[0034] like Figure 2 As shown, the data anomaly alarm method provided by the embodiment of the present invention includes the following S201-S204.

[0035] S201. The alarm device obtains fluctuation data of a target currency pair within a first time period.

[0036] The fluctuation data within a time period includes a time identifier for reflecting whether the time period is within working hours, and the fluctuation data within a time period also includes the number of received updates and the number of sent updates of the target currency pair, and / or the received volatility and the sent volatility of the target currency pair.

[0037] As a possible implementation method, the alarm device may obtain from the server the time identifier corresponding to the target currency pair in the first time period, as well as the number of updates received and sent by the target currency pair in the first time period.

[0038] It should be noted that the time flag for the target currency pair within a time period reflects whether the foreign exchange market or the local market is a holiday during that time period. For example, if either the foreign exchange market or the local market of the target currency pair is a holiday, the time flag is "1"; if both the foreign exchange market and the local market of the target currency pair are business days, the time flag is "0".

[0039] The number of received updates indicates the number of price updates of the target currency pair received by the server from the external interface within a time period, and the number of sent updates indicates the number of times the server sent the price of the target currency pair to the electronic quote device within a time period. Simultaneously, the alarm device obtains the received price and sent price of the target currency pair within the first time period from the server, and determines the received volatility of the target currency pair within the first time period based on the obtained received price, and determines the sent volatility of the target currency pair within the first time period based on the obtained sent price.

[0040] The received volatility is used to indicate the price fluctuation of the target currency pair received by the server from the external interface within a time period, and the sent volatility is used to indicate the price fluctuation of the target currency pair sent by the server to the electronic quoter within a time period.

[0041] The following shows one implementation of determining the received volatility of a target currency pair over a time period:

[0042] The alarm device may calculate the downward volatility of the target currency pair within a time period based on the received price of the target currency pair received within the time period and the following formula 1:

[0043]

[0044] Among them, V s is the received volatility of the target currency pair in a period of time, P i is the price of the target currency pair received for the i-th time in a time period; Pi+1 is the price of the target currency pair received for the i+1-th time in a time period, and n represents the number of times the price of the target currency pair is received in a time period.

[0045] Similarly, the alarm device can also use Formula 2, which is similar to Formula 1 above, to calculate the downward volatility of the target currency pair within a time period:

[0046]

[0047] Among them, V Q is the volatility of the target currency pair within a period of time, Q i is the price of the target currency pair issued for the i-th time within a time period; Qi+1 is the price of the target currency pair issued the i+1th time within a time period, and m represents the number of times the price of the target currency pair is issued within a time period. As another implementation method for obtaining the received volatility and the issued volatility of the target currency pair within the first time period, the alarm device can directly obtain the received volatility and the issued volatility of the target currency pair within the first time period from the server.

[0048] The specific implementation of the server determining the received volatility and issuing the volatility of the currency pair within a time period can be referred to the above description of the embodiment of the present invention, which will not be repeated here. The difference lies in the different execution entities.

[0049] In the first case, the fluctuation data within a time period obtained by the alarm device from the server includes a time identifier, a number of received updates, and a number of sent updates of the target currency pair within the time period.

[0050] In the second case, the volatility data within a time period acquired by the alarm device from the server includes a time identifier, a received volatility rate, and a sent volatility rate of the target currency pair within the time period.

[0051] In the third case, the volatility data within a time period obtained by the alarm device from the server includes a time identifier, a number of received updates, a number of sent updates, a received volatility, and a sent volatility of the target currency pair within the time period.

[0052] Optionally, the volatility data may also include information such as the identifier of the target currency pair, the length of the time period, the trading volume, and the median price.

[0053] It should be noted that the above data is stored in the database in comma-separated values (CSV) file format by date, for example, 2019-08-22-Update Count-014001.csv, where 2019-08-22 represents the date, Update Count represents the number of updates issued, and 014001 represents the currency pair identifier.

[0054] S202: The alarm device determines target parameters based on the fluctuation data of the target currency pair in the first time period and a preset neural network model.

[0055] The target parameter is used to reflect the number of updates and / or volatility within a time period.

[0056] As a possible implementation method, the alarm device inputs the fluctuation data of the target currency pair in the first time period into a preset neural network model, and uses the data output by the neural network model as the target parameter.

[0057] As another possible implementation method, the alarm device preprocesses the fluctuation data of the target currency pair within the first time period, inputs the preprocessed fluctuation data into a preset neural network model, and uses the data output by the neural network model as the target parameter.

[0058] It should be noted that the neural network model can be a neural network based on unsupervised learning, capable of performing convergence and clustering processing on the input data. The preprocessing can be a discretization operation. By discretizing the fluctuating data, the raw data can be converted into input data that meets the requirements of the neural network model and can also effectively reduce data noise caused by accidental factors.

[0059] In the first case, the target parameter reflects the number of updates of the target currency pair during the first time period. In the second case, the target parameter reflects the volatility of the target currency pair during the first time period. In the third case, the target parameter reflects both the number of updates and the volatility of the target currency pair during the first time period.

[0060] The target parameter can be one parameter or multiple parameters. The target parameter can be expressed in the form of a matrix, which is not specifically limited here.

[0061] The specific implementation of this step can refer to the subsequent description of the embodiment of the present invention and will not be repeated here.

[0062] S203: The alarm device determines whether the target parameter meets the alarm condition.

[0063] The alarm condition is determined based on a neural network model and fluctuation data of a target currency pair in a plurality of second time periods, where the plurality of second time periods include a time period before the first time period.

[0064] It should be noted that the alarm condition is composed of target parameters within multiple second time periods. The alarm condition can be a value, a range, multiple ranges, a two-dimensional area, or a three-dimensional space, and is not specifically limited here.

[0065] For example, when the number of parameters output by the neural network is one, the alarm condition can be a numerical value or a range. When the number of parameters output by the neural network is multiple, the alarm condition can be multiple ranges, and the number of multiple ranges is the same as the number of parameters.

[0066] For the specific implementation of this step, please refer to the subsequent description of the embodiment of the present invention and will not be repeated here.

[0067] S204: If the target parameter meets the alarm condition, the alarm device generates alarm information.

[0068] As a possible implementation manner, the alarm device generates alarm information after determining that the target parameter meets the alarm condition.

[0069] Subsequently, the alarm device sends alarm information to the server, so that the server issues an alarm after receiving the alarm information.

[0070] It should be noted that the alarm information is used to indicate data anomalies, and the alarm information may include the name of the target currency pair and the duration of the first time period.

[0071] In one design, the neural network model provided by the embodiment of the present invention may be a long short-term memory (LSTM) neural network model. Figure 3 A structural diagram of the LSTM neural network model is provided. Figure 3 As shown, the LSTM neural network model includes a loop structure, which is used to cyclically process fluctuation data in different time periods according to the long-term memory characteristics of the LSTM neural network model, and the loop structure includes multiple LSTM units.

[0072] Figure 4 A structural diagram of an LSTM unit is provided. Figure 4 As shown in Figure 1, the LSTM unit includes a forget gate, an input gate, and an output gate.

[0073] The forget gate of the LSTM unit is used to determine the information that needs to be discarded from the previous LSTM unit through the forget gate parameters.

[0074] Exemplarily, the forget gate parameters satisfy the following formula 3:

[0075] f τ =σ(W f ·[h τ-1 ,x τ ]+b f )Formula 3

[0076] Among them, f τ is the forget gate parameter of the LSTM unit in the τth time period (the size is between 0-1), σ is the sigmoid function, W f is the weight matrix of the forget gate parameter of the LSTM unit, b f is the bias term of the forget gate parameter of the LSTM unit, h τ-1 is the output value of the LSTM unit in the τ-1th time period, x τ is the volatility data of the target currency pair in the τth time block.

[0077] The input gate of the LSTM unit is used to determine the information added to the LSTM unit through the input gate parameters. The input gate parameters consist of two parts: one part is composed of the sigmoid function σ, which is used to determine the information that needs to be updated; the other part is composed of the tanh function, which is used to determine the alternative update information.

[0078] For example, the input gate parameters satisfy the following formulas 4 and 5:

[0079] i τ =σ(W i ·[h τ-1 ,x τ ]+b i ) Formula 4

[0080]

[0081] Among them, i τ and are the two parts of the input gate parameters of the LSTM unit in the τth time period; W i is the input gate parameter i of the LSTM unit τ The weight matrix, b i is the input gate parameter i of the LSTM unit τ The bias term, W C is the input gate parameter of the LSTM unit The weight matrix, b C is the input gate parameter of the LSTM unit The bias term.

[0082] The cell state within a time period output by the LSTM unit is determined by the forget gate parameters and the input gate parameters;

[0083] For example, the cell state within a time period output by the LSTM unit satisfies the following formula 6:

[0084]

[0085] C τ is the cell state of the LSTM unit in the τth time period; C τ-1 is the cell state of the LSTM unit in the τ-1th time period.

[0086] The output gate of the LSTM unit determines the output of the LSTM unit through the output gate parameter, which is also composed of a sigmoid function;

[0087] Exemplarily, the output gate parameters satisfy the following formula 7:

[0088] o τ =σ(W o [hτ-1 ,x τ ]+b o ) Formula 7

[0089] o τ is the output gate parameter of the LSTM unit in the τth time period, W o is the weight matrix of the output gate parameters of the LSTM unit, b o is the bias term of the output gate parameter of the LSTM unit.

[0090] The final output value of the LSTM unit is determined by the output gate and the cell state. The final output value satisfies the following formula 8:

[0091] h τ =o τ *tanh(C τ ) Formula 8

[0092] h τ is the output value of the LSTM unit in the τth time period, which is used to represent the output hidden state in the τth time period.

[0093] In the loop structure of the LSTM neural network model, the initial cell state C1 is a set of random values (generally a zero matrix), x1 is the fluctuation data in the initial time period, and the initial output value h0 of the LSTM unit is a zero matrix.

[0094] For example, according to x1 and h0, f1, i1, and and o1;

[0095] According to o1 and C1, the output value h1 of the initial LSTM unit can be calculated according to the preset output formula.

[0096] According to x2 and h1, we can calculate f2, i2, and and o2;

[0097] According to C1, f2, i2 and C2 can be calculated based on the cell state parameter formula;

[0098] Based on o2 and C2, h2 can be calculated according to the preset output formula.

[0099] According to x3 and h2, we can calculate f3, i3, and and o3;

[0100] According to C2, f3, i3 and C3 can be calculated based on the cell state parameter formula;

[0101] Based on o3 and C3, h3 can be calculated according to the preset output formula.

[0102] Sequentially, referring to the above calculation process, the output value of the LSTM unit in the next time period can be calculated from the output gate parameters and cell state parameters of the LSTM unit in the previous time period. Therefore, according to the initial output value h0 of the loop structure of the LSTM neural network model, the output value of the preset number of time periods before the current time period can be obtained by continuous iterative calculation. Among them, the output value of the time period before the current time period can be expressed as h τ-1 .

[0103] In the above calculation process, x1, x2, ..., x τ-1 The alarm device can obtain the corresponding fluctuation data in each time period from the server.

[0104] In the case where the above-mentioned preset neural network model is the above-mentioned LSTM neural network model, in order to be able to determine the alarm conditions, such as Figure 5 As shown, the data anomaly alarm method provided by the embodiment of the present invention further includes the following S301-S303 before the above S203.

[0105] S301. The alarm device obtains fluctuation data of a target currency pair in multiple second time periods.

[0106] The specific implementation of this step can refer to the specific description of S201 in the embodiment of the present invention, which will not be repeated here. The difference is that the acquired fluctuation data belongs to a different time period.

[0107] At the same time, the content of the fluctuation data within a time period acquired by the alarm device in this step can refer to the description of the three situations in S201.

[0108] S302. The alarm device obtains multiple historical parameters according to the LSTM neural network and the fluctuation data in multiple second time periods.

[0109] Wherein, one historical parameter corresponds to the fluctuation data in a second time period, and one historical parameter includes a historical cell state and a historical output hidden state.

[0110] The specific implementation of this step can refer to the specific description of S202 in the embodiment of the present invention, which will not be repeated here. The difference is that the acquired fluctuation data belongs to different time periods, and the preset neural network model is an LSTM neural network.

[0111] It should be noted that when the alarm device inputs fluctuation data within a time period, the LSTM neural network model outputs a historical cell state C τ and a historical output hidden state h τ .

[0112] If the fluctuation data of a time period is input as the first case in S201, the historical cell state C τ Used to indicate the number of updates that a currency pair needs to remember within a period of time, and the historical output hidden state h τ Used to indicate the number of hidden updates a currency pair needs to undergo within a time period.

[0113] If the fluctuation data of a time period is input as the second case in S201, the historical cell state C τ Used to represent the volatility of a currency pair that needs to be remembered within a period of time, the historical output hidden state h τ Used to indicate the hidden volatility of a currency pair over a period of time.

[0114] If the fluctuation data of a time period is input as the third case in S201, the historical cell state C τ Used to indicate the number of updates and volatility that a currency pair needs to remember within a period of time, and the historical output hidden state h τ Used to indicate the number of hidden updates and volatility of a currency pair within a time period.

[0115] S303: The alarm device determines an alarm condition according to the multiple historical cell states and the multiple historical output hidden states included in the multiple historical parameters.

[0116] As a possible implementation method, the alarm device determines a cell state range based on multiple historical cell states, and determines an output hidden state range based on multiple historical output hidden states, and determines the cell state range and the output hidden state range as alarm conditions.

[0117] The specific implementation of this step can refer to the subsequent description of the embodiment of the present invention and will not be repeated here.

[0118] In one design, Figure 6 As shown, the alarm conditions provided by the embodiment of the present invention include the cell state range and the output hidden state range. In this case, in order to determine the alarm conditions, the above S303 provided by the embodiment of the present invention may specifically include the following S3031-S3032.

[0119] S3031. The alarm device determines a cell state range based on multiple historical cell states.

[0120] The maximum value of the cell state range is greater than the maximum value of the multiple historical cell states, and the minimum value of the cell state range is less than the minimum value of the multiple historical cell states.

[0121] As a possible implementation method, the alarm device determines a historical cell state maximum value from multiple historical cell states, and determines the sum of the historical cell state maximum value and the first threshold as the maximum value of the cell state range;

[0122] At the same time, the alarm device determines a historical cell state minimum value from multiple historical cell states, and determines the difference between the historical cell state minimum value and the second threshold value as the minimum value of the cell state range.

[0123] It should be noted that the first threshold and the second threshold are both values greater than 0. The first threshold and the second threshold can be set in advance by the operation and maintenance personnel in the alarm device. In actual applications, the first threshold and the second threshold can be the same or different, and the embodiment of the present invention does not limit this.

[0124] S3032. The alarm device determines an output hidden state range based on multiple historical output hidden states.

[0125] The maximum value of the output hidden state range is greater than the maximum values of the multiple historical output hidden states, and the minimum value of the output hidden state range is less than the minimum value of the multiple historical output hidden states.

[0126] As a possible implementation manner, the alarm device determines an output hidden state maximum value from multiple historical output hidden states, and determines the sum of the output hidden state maximum value and a third threshold as the maximum value of the output hidden state range.

[0127] At the same time, the alarm device determines a historical output hidden state minimum value from multiple historical output hidden states, and determines the difference between the historical output hidden state minimum value and a fourth threshold as the minimum value of the output hidden state range.

[0128] It should be noted that the third threshold and the fourth threshold are both values greater than 0. The third threshold and the fourth threshold can be pre-set in the alarm device by the operation and maintenance personnel. In actual applications, the third threshold and the fourth threshold can be the same or different, and the embodiment of the present invention does not limit this.

[0129] In one design, the target parameters provided by the embodiment of the present invention include the target cell state and the target output hidden state. In this case, in order to determine whether the target parameters meet the alarm conditions, such as Figure 7 As shown, the above S203 provided in the embodiment of the present invention specifically includes the following S2031-S2032.

[0130] S2031. The alarm device determines whether the target cell state exceeds the cell state range and whether the target output hidden state exceeds the output hidden state range.

[0131] In practical applications, we can first determine whether the target cell state exceeds the cell state range, and if the target cell state does not exceed the cell state range, determine whether the target output hidden state exceeds the output hidden state range; or we can first determine whether the target output hidden state exceeds the output hidden state range, and if the target output hidden state does not exceed the output hidden state range, determine whether the target cell state exceeds the cell state range.

[0132] S2032: If the target cell state exceeds the cell state range, or the target output hidden state exceeds the output hidden state range, the alarm device determines that the target parameter meets the alarm condition.

[0133] Exemplarily, when the target cell state is less than the minimum value of the cell state range, the alarm device determines that the target parameter meets the alarm condition.

[0134] In actual applications, S2031 may be executed first and then S2032, or S2032 may be executed first and then S2031, or S2031 and S2032 may be executed simultaneously, which is not limited in the embodiment of the present invention.

[0135] In one design, the alarm condition provided by the embodiment of the present invention may also include an alarm area. In this case, in order to determine the alarm condition, such as Figure 8 As shown, the above S303 provided by the embodiment of the present invention may specifically include the following S3033-S3034.

[0136] S3033. For the first historical parameter, the alarm device determines a historical alarm position corresponding to the first historical parameter according to the first historical cell state and the first historical output hidden state included in the first historical parameter.

[0137] The first historical parameter is any one of a plurality of historical parameters, and the first historical cell state and the first historical output hidden state are respectively coordinate values of the historical alarm position corresponding to the first historical parameter.

[0138] As a possible implementation method, the alarm device determines the first historical cell state and the first historical output hidden state as the coordinate values of the historical alarm position corresponding to the first historical parameter, so as to obtain the historical alarm position corresponding to the first historical parameter.

[0139] It should be noted that when the first historical cell state is the x-axis coordinate value, the first historical output hidden state is the y-axis coordinate value; or, when the first historical output hidden state is the x-axis coordinate value, the first historical cell state is the y-axis coordinate value. This embodiment does not limit this.

[0140] It can be understood that, through S3033, the alarm device can determine the historical alarm positions corresponding to the above multiple historical parameters.

[0141] S3034. The alarm device generates an alarm area according to the historical alarm positions corresponding to the multiple historical parameters.

[0142] As a possible implementation method, the alarm device constructs a two-dimensional coordinate system, and locates the historical alarm positions corresponding to multiple historical parameters in the two-dimensional coordinate system through a preset algorithm, and uses the area formed after positioning as the alarm area.

[0143] Optionally, the warning area may be a regular pattern or an irregular pattern, which is not limited in the embodiment of the present invention.

[0144] Illustratively, in a two-dimensional coordinate system, the alarm area may be the largest area formed by interconnecting historical alarm positions corresponding to multiple historical parameters, or may be the largest circumscribed circle formed by historical alarm positions corresponding to multiple historical parameters.

[0145] Optionally, the preset algorithm may be a binary information gain optimization algorithm (BINGO).

[0146] In one design, the target parameters provided by the embodiment of the present invention include the target cell state and the target output hidden state. In this case, in order to determine whether the target parameters meet the alarm conditions, such as Figure 9 As shown, the above S203 provided in the embodiment of the present invention specifically includes the following S2033-S2035.

[0147] S2033. The alarm device determines the target alarm position corresponding to the target parameter according to the target cell state and the target output hidden state.

[0148] Among them, the target cell state and the target output hidden state are the coordinate values of the target alarm position respectively.

[0149] The specific implementation of this step can refer to the specific description of S3033 in the embodiment of the present invention, which will not be repeated here. The difference lies in that the alarm location generated is different, and the parameters required to generate the alarm location are different.

[0150] S2034: The alarm device determines whether the minimum distance between the target alarm position and the alarm area is greater than a fifth threshold.

[0151] A specific example of determining the minimum distance is shown below:

[0152] The warning device emits rays from the target warning position toward the warning area, forming multiple line segments. Each line segment corresponds to a ray, and the line segment is a line segment between the target warning position and a target intersection point on the ray. The target intersection point is the intersection point closest to the target warning position among at least one intersection point of the ray with the warning area. Furthermore, the warning device determines the minimum distance as the value with the smallest length among the multiple line segments.

[0153] S2035: If the minimum distance between the target alarm position and the alarm area is greater than a fifth threshold, the alarm device determines that the target parameter meets the alarm condition.

[0154] It should be noted that the fifth threshold is a value greater than 0, and the fifth threshold can be set in advance in the alarm device by the operation and maintenance personnel.

[0155] In one design, in order to further ensure the accuracy of the alarm results, such as Figure 2 As shown, the above-mentioned alarm method provided by the embodiment of the present invention further includes the following S205.

[0156] S205: If the target parameter does not meet the alarm condition, the alarm device updates the alarm condition according to the target parameter.

[0157] As a possible implementation method, when the alarm condition includes a cell state range and an output hidden state range, the alarm device updates the target cell state corresponding to the target parameter to the above-mentioned multiple historical cell states, and executes the above-mentioned step S3031, and updates the target output hidden state to multiple historical output hidden states, and executes the above-mentioned step S3032.

[0158] As another possible implementation method, when the alarm condition includes an alarm area, the alarm device updates the target alarm position to a historical alarm position corresponding to a plurality of historical parameters, and executes the above step S3034.

[0159] Embodiments of the present invention provide a data anomaly alarm method and device. First, an alarm device obtains fluctuation data for a target currency pair within a first time period. Based on the obtained fluctuation data and a preset neural network model, the device determines a target parameter reflecting the number of updates and / or volatility of the target currency pair within the time period. Because the fluctuation data within a time period includes a time stamp indicating whether the time period falls within working hours, and also includes the number of received and issued updates for the target currency pair, and / or the received and issued volatility of the target currency pair, compared to existing technologies, the device can utilize multi-dimensional information to determine the target parameter, which more comprehensively reflects the fluctuations of the target currency pair within the first time period. Furthermore, because the alarm device uses alarm conditions as the basis for determining whether to issue an alarm, and these alarm conditions are determined by the alarm device based on fluctuation data from multiple historical time periods under the same conditions, the determination criteria can be applied to the target parameters. Consequently, when determining whether to issue an alarm based on the target parameters and alarm conditions, the alarm result can be more accurate.

[0160] The above mainly introduces the solution provided by the embodiment of the present invention from the perspective of method. In order to realize the above functions, it includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiment disclosed herein, the embodiment of the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0161] In the embodiments of the present invention, the functional modules of the device described above can be divided according to the above method examples. For example, each functional module can be divided according to each function, or two or more functions can be integrated into a single processing module. The integrated modules can be implemented in the form of hardware or software functional modules. Optionally, the module division in the embodiments of the present invention is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used.

[0162] Figure 10 This is a schematic diagram of the structure of an alarm device provided by an embodiment of the present invention. Figure 10 As shown, the alarm device 40 is used to generate alarm information when the target parameter meets the alarm condition, for example, Figure 2The alarm method shown in FIG. 4 shows an alarm device 40 including an acquisition unit 401 , a determination unit 402 and a generation unit 403 .

[0163] The acquisition unit 401 is configured to acquire the volatility data of the target currency pair within a first time period; the volatility data within a time period includes a time mark indicating whether the time period is within working hours, and the volatility data within a time period also includes the number of received updates and the number of issued updates of the target currency pair, and / or the received volatility rate and the issued volatility rate of the target currency pair. For example, Figure 2 As shown, the acquiring unit 401 can be used to execute S201.

[0164] The determination unit 402 is used to determine the target parameter based on the preset neural network model and the volatility data of the target currency pair in the first time period obtained by the acquisition unit 401; the target parameter is used to reflect the number of updates and / or volatility in a time period. For example, Figure 2 As shown, the determining unit 402 can be used to execute S202.

[0165] The generating unit 403 is configured to generate an alarm message if the target parameter determined by the determining unit satisfies an alarm condition; the alarm condition is determined based on the neural network model and the fluctuation data of the target currency pair in multiple second time periods, and the multiple second time periods include the time period before the first time period. For example, Figure 2 As shown, the generating unit 403 can be used to execute S204.

[0166] Optionally, the neural network model provided in the embodiment of the present invention is a long short-term memory LSTM neural network model. In this case, Figure 10 As shown, the acquisition unit 401 provided in the embodiment of the present invention is also used to acquire the fluctuation data of the target currency pair in multiple second time periods. Figure 5 As shown, the acquiring unit 401 can be used to execute S301.

[0167] The determination unit 402 is further configured to obtain a plurality of historical parameters according to the LSTM neural network and the fluctuation data in the second time period obtained by the acquisition unit; one historical parameter corresponds to one fluctuation data in the second time period, and one historical parameter includes a historical cell state and a historical output hidden state. For example, Figure 5 As shown, the determining unit 402 can be used to execute S302.

[0168] The determination unit 402 is further configured to determine an alarm condition based on the multiple historical cell states and multiple historical output hidden states included in the multiple historical parameters determined by the determination unit. Figure 5 As shown, the determining unit 402 can be used to execute S303.

[0169] Optionally, the alarm conditions provided by the embodiment of the present invention include the cell state range and the output hidden state range. In this case, if Figure 10 As shown, the determining unit 402 provided in the embodiment of the present invention is specifically used to determine a cell state range based on multiple historical cell states; the maximum value of the cell state range is greater than the maximum value of the multiple historical cell states, and the minimum value of the cell state range is less than the minimum value of the multiple historical cell states. Figure 6 As shown, the determining unit 402 can be used to execute S3031.

[0170] The determining unit 402 is specifically configured to determine an output hidden state range based on multiple historical output hidden states; the maximum value of the output hidden state range is greater than the maximum values of the multiple historical output hidden states, and the minimum value of the output hidden state range is less than the minimum value of the multiple historical output hidden states. Figure 6 As shown, the determining unit 402 can be used to execute S3032.

[0171] Optionally, the target parameters provided by the embodiment of the present invention include the target cell state and the target output hidden state. In this case, Figure 10 As shown, the alarm device 40 provided by the embodiment of the present invention further includes a judgment unit 404. The judgment unit 404 is specifically configured to determine that the target parameter meets the alarm condition if the target cell state exceeds the cell state range, or the target output hidden state exceeds the output hidden state range. Figure 7 As shown, the judgment unit 404 can be used to execute S2032.

[0172] Optionally, the alarm condition provided by the embodiment of the present invention may also include an alarm area. In this case, if Figure 10 As shown, the determining unit 402 provided in this embodiment of the present invention is specifically configured to:

[0173] For a first historical parameter, the historical alarm position corresponding to the first historical parameter is determined based on the first historical cell state and the first historical output hidden state included in the first historical parameter; the first historical parameter is any one of multiple historical parameters, and the first historical cell state and the first historical output hidden state are respectively the coordinate values of the historical alarm position corresponding to the first historical parameter. For example, Figure 8 As shown, the determining unit 402 can be used to execute S3033.

[0174] Generate an alarm area based on the historical alarm locations corresponding to multiple historical parameters. For example, Figure 8 As shown, the determining unit 402 can be used to execute S3034.

[0175] Optionally, the target parameters provided by the embodiment of the present invention include the target cell state and the target output hidden state. In this case, Figure 10 As shown, the alarm device 40 provided in the embodiment of the present invention further includes a judgment unit 404. The judgment unit 404 is specifically configured to:

[0176] According to the target cell state and the target output hidden state, the target alarm position corresponding to the target parameter is determined; the target cell state and the target output hidden state are the coordinate values of the target alarm position respectively. Figure 9 As shown, the judgment unit 404 can be used to execute S2033.

[0177] If the minimum distance between the target alarm position and the alarm area is greater than the fifth threshold, it is determined that the target parameter meets the alarm condition. Figure 9 As shown, the judgment unit 404 can be used to execute S2035.

[0178] Optional, such as Figure 10 As shown, the alarm device 40 further includes an updating unit 405. The updating unit 405 is configured to update the alarm condition according to the target parameter if the target parameter does not meet the alarm condition. Figure 2 As shown, the updating unit 405 can be used to execute S205.

[0179] In the case of implementing the functions of the above-mentioned integrated modules in the form of hardware, the embodiment of the present invention provides a possible structural diagram of the alarm device involved in the above-mentioned embodiment. Figure 11 As shown, an alarm device 50 is used to solve the problem of inaccurate alarms, for example, for executing Figure 2 The alarm method for data anomaly shown in FIG. 1 includes a processor 501 , a memory 502 and a bus 503 . The processor 501 and the memory 502 may be connected via the bus 503 .

[0180] Processor 501 is the control center of alarm device 50 and can be a single processor or a collective term for multiple processing elements. For example, processor 501 can be a general-purpose central processing unit (CPU) or other general-purpose processor. The general-purpose processor can be a microprocessor or any conventional processor.

[0181] As an embodiment, the processor 501 may include one or more CPUs, such as Figure 11 CPU 0 and CPU 1 are shown in Figure 1.

[0182] The memory 502 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, an electrically erasable programmable read-only memory (EEPROM), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0183] As a possible implementation, the memory 502 can exist independently of the processor 501 and can be connected to the processor 501 via a bus 503 to store instructions or program codes. When the processor 501 calls and executes the instructions or program codes stored in the memory 502, the alarm method provided in the embodiment of the present invention can be implemented.

[0184] In another possible implementation, the memory 502 may also be integrated with the processor 501 .

[0185] Bus 503 can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. This bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0186] It should be pointed out that Figure 11 The structure shown does not constitute a limitation of the alarm device 50. Figure 11 In addition to the components shown, the warning device 50 may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0187] As an example, combining Figure 10 The functions implemented by the acquisition unit 401, the determination unit 402, the generation unit 403, the judgment unit 404 and the update unit 405 in the alarm device 40 are the same as those implemented by the Figure 11 The function of the processor 501 in is the same.

[0188] Optional, such as Figure 11 As shown, the alarm device 50 provided by the embodiment of the present invention may further include a communication interface 504 .

[0189] The communication interface 504 is used to connect to other devices via a communication network. The communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc. The communication interface 504 can include a receiving unit for receiving data and a sending unit for sending data.

[0190] In one design, in the alarm device provided by an embodiment of the present invention, the communication interface can also be integrated into the processor.

[0191] Figure 12 FIG. 2 shows another hardware structure of the alarm device in an embodiment of the present invention. Figure 12 As shown, the alarm device 60 may include a processor 601 and a communication interface 602. The processor 601 is coupled to the communication interface 602.

[0192] The functions of the processor 601 may refer to the description of the processor 501. In addition, the processor 601 also has a storage function, which may refer to the function of the memory 502.

[0193] The communication interface 602 is used to provide data to the processor 601. The communication interface 602 can be an internal interface of the communication device, or an external interface of the communication device (equivalent to the communication interface 504).

[0194] It should be pointed out that Figure 12 The structure shown in does not constitute a limitation on the alarm device 60, except Figure 12 In addition to the components shown, the warning device 60 may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0195] Through the description of the above embodiments, those skilled in the art will clearly understand that for the sake of convenience and brevity, only the division of the above-mentioned functional units is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units as needed, that is, the internal structure of the device can be divided into different functional units to complete all or part of the functions described above. The specific working processes of the above-mentioned systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0196] An embodiment of the present invention further provides a computer-readable storage medium, in which instructions are stored. When a computer executes the instructions, the computer executes each step in the method flow shown in the above method embodiment.

[0197] An embodiment of the present invention provides a computer program product including instructions. When the instructions are executed on a computer, the computer is caused to execute the data anomaly alarm method in the above method embodiment.

[0198] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk. Random Access Memory (RAM), Read-Only Memory (ROM), Erasable Programmable Read Only Memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any other form of computer-readable storage medium in a suitable combination of the above, or a numerical value in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In embodiments of the present invention, computer-readable storage media may be any tangible media that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0199] Since the alarm device, computer-readable storage medium, and computer program product in the embodiments of the present invention can be applied to the above method, the technical effects that can be obtained can also refer to the above method embodiments, and the embodiments of the present invention will not be repeated here.

[0200] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or replacements within the technical scope disclosed by the present invention should be included in the protection scope of the present invention.

Claims

1. A data anomaly alarm method, characterized in that: include: Get the volatility data of the target currency pair in the first time period; The fluctuation data within a time period includes a time identifier indicating whether the time period falls within working hours, and the fluctuation data within the time period also includes the number of received updates and the number of issued updates for the target currency pair, and / or the received volatility and the issued volatility of the target currency pair; determining target parameters based on the volatility data of the target currency pair during the first time period and a preset neural network model; The target parameter is used to reflect the number of updates and / or fluctuation rate within the time period; If the target parameter meets the alarm condition, an alarm message is generated; The alarm condition is determined based on the neural network model and fluctuation data of the target currency pair in multiple second time periods, wherein the multiple second time periods include a time period before the first time period; The neural network model is a long short-term memory (LSTM) neural network model, and the method further includes: Obtaining fluctuation data of the target currency pair within the plurality of second time periods; Obtaining a plurality of historical parameters based on the LSTM neural network and the fluctuation data in the plurality of second time periods; one historical parameter corresponds to one fluctuation data in the second time period, and the one historical parameter includes a historical cell state and a historical output hidden state; Determining the alarm condition according to a plurality of historical cell states and a plurality of historical output hidden states included in the plurality of historical parameters; The alarm condition includes an alarm area, and determining the alarm condition according to a plurality of historical cell states and a plurality of historical output hidden states included in the plurality of historical parameters includes: For a first historical parameter, determining a historical alarm position corresponding to the first historical parameter based on a first historical cell state and a first historical output hidden state included in the first historical parameter; the first historical parameter is any one of the multiple historical parameters, and the first historical cell state and the first historical output hidden state are respectively coordinate values of the historical alarm position corresponding to the first historical parameter; generating the alarm area according to the historical alarm positions corresponding to the multiple historical parameters; The target parameters include a target cell state and a target output hidden state, and the method further includes: Determining a target alarm position corresponding to the target parameter according to the target cell state and the target output hidden state; the target cell state and the target output hidden state are respectively coordinate values of the target alarm position; If the minimum distance between the target alarm position and the alarm area is greater than a fifth threshold, determining that the target parameter meets the alarm condition; Among them, the cell state within a time period output by the LSTM unit satisfies the following formula: C τ is the cell state of the LSTM unit in the τth time period; C τ-1 is the cell state of the LSTM unit in the τ-1th time period, i τ and are the two parts of the input gate parameters of the LSTM unit in the τth time period, f τ is the forget gate parameter of the LSTM unit in the τth time period.

2. The data anomaly alarm method according to claim 1, characterized in that: The alarm condition includes a cell state range and an output hidden state range, and determining the alarm condition based on a plurality of historical cell states and a plurality of historical output hidden states included in the plurality of historical parameters includes: Determining a cell state range based on the multiple historical cell states; wherein a maximum value of the cell state range is greater than a maximum value among the multiple historical cell states, and a minimum value of the cell state range is less than a minimum value among the multiple historical cell states; The output hidden state range is determined based on the multiple historical output hidden states; the maximum value of the output hidden state range is greater than the maximum value of the multiple historical output hidden states, and the minimum value of the output hidden state range is less than the minimum value of the multiple historical output hidden states.

3. The data anomaly alarm method according to claim 2, characterized in that: The target parameters include target cell state and target output hidden state; The method further comprises: If the target cell state exceeds the cell state range, or the target output hidden state exceeds the output hidden state range, it is determined that the target parameter meets the alarm condition.

4. The data anomaly alarm method according to any one of claims 1 to 3, characterized in that: The method further comprises: If the target parameter does not meet the alarm condition, the alarm condition is updated according to the target parameter.

5. An alarm device, characterized in that: It includes an acquisition unit, a determination unit, and a generation unit; The acquisition unit is configured to acquire fluctuation data of a target currency pair within a first time period; the fluctuation data within a time period includes a time identifier indicating whether the time period falls within working hours, and the fluctuation data within a time period also includes a number of received updates and a number of issued updates for the target currency pair, and / or a received volatility and a issued volatility for the target currency pair; The determining unit is configured to determine a target parameter based on a preset neural network model and the fluctuation data of the target currency pair within the first time period acquired by the acquiring unit; The target parameter is used to reflect the number of updates and / or fluctuation rate within the time period; The generating unit is configured to generate alarm information if the target parameter determined by the determining unit meets an alarm condition; The alarm condition is determined based on the neural network model and fluctuation data of the target currency pair in multiple second time periods, wherein the multiple second time periods include a time period before the first time period; The neural network model is a long short-term memory (LSTM) neural network model. The acquisition unit is further configured to acquire fluctuation data of the target currency pair within the plurality of second time periods; The determining unit is further configured to obtain a plurality of historical parameters based on the LSTM neural network and the fluctuation data in the plurality of second time periods obtained by the obtaining unit; one historical parameter corresponds to one fluctuation data in the second time period, and the one historical parameter includes a historical cell state and a historical output hidden state; The determining unit is further configured to determine the alarm condition according to a plurality of historical cell states and a plurality of historical output hidden states included in the plurality of historical parameters determined by the determining unit; The alarm condition includes an alarm area; and the determining unit is specifically configured to: For a first historical parameter, determining a historical alarm position corresponding to the first historical parameter based on a first historical cell state and a first historical output hidden state included in the first historical parameter; the first historical parameter is any one of the multiple historical parameters, and the first historical cell state and the first historical output hidden state are respectively coordinate values of the historical alarm position corresponding to the first historical parameter; generating the alarm area according to the historical alarm positions corresponding to the multiple historical parameters; The target parameters include a target cell state and a target output hidden state; the alarm device further includes a judgment unit; the judgment unit is specifically configured to: Determining a target alarm position corresponding to the target parameter according to the target cell state and the target output hidden state; the target cell state and the target output hidden state are respectively coordinate values of the target alarm position; If the minimum distance between the target alarm position and the alarm area is greater than a fifth threshold, determining that the target parameter meets the alarm condition; Among them, the cell state within a time period output by the LSTM unit satisfies the following formula: C τ is the cell state of the LSTM unit in the τth time period; C τ-1 is the cell state of the LSTM unit in the τ-1th time period, i τ and are the two parts of the input gate parameters of the LSTM unit in the τth time period, f τ is the forget gate parameter of the LSTM unit in the τth time period.

6. The alarm device according to claim 5, characterized in that: The alarm condition includes a cell state range and an output hidden state range; the determination unit is specifically configured to: Determining a cell state range based on the multiple historical cell states; wherein a maximum value of the cell state range is greater than a maximum value among the multiple historical cell states, and a minimum value of the cell state range is less than a minimum value among the multiple historical cell states; The output hidden state range is determined based on the multiple historical output hidden states; the maximum value of the output hidden state range is greater than the maximum value of the multiple historical output hidden states, and the minimum value of the output hidden state range is less than the minimum value of the multiple historical output hidden states.

7. The alarm device according to claim 6, characterized in that: The target parameters include target cell state and target output hidden state; the alarm device also includes a judgment unit; The judgment unit is specifically configured to determine that the target parameter meets the alarm condition if the target cell state exceeds the cell state range, or the target output hidden state exceeds the output hidden state range.

8. The alarm device according to any one of claims 5 to 7, characterized in that: The alarm device further includes an updating unit; The updating unit is configured to update the alarm condition according to the target parameter if the target parameter does not meet the alarm condition.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions, which, when executed by a computer, enable the computer to execute the data anomaly alarm method according to any one of claims 1 to 4.

10. An alarm device, characterized in that: include: A processor and a memory; wherein the memory is used to store one or more programs, and the one or more programs include computer-executable instructions. When the alarm device is running, the processor executes the computer-executable instructions stored in the memory to enable the alarm device to execute the data anomaly alarm method described in any one of claims 1 to 4.

Images