CN111863276B - Hand, foot and mouth disease prediction method, electronic equipment and medium using fine-grained data - Google Patents

Abstract

The present application provides a hand, foot and mouth disease prediction method and device using fine-grained data, electronic equipment, and a computer-readable medium. The method includes: obtaining historical case data of hand, foot and mouth disease; preprocessing the historical case data, and counting them as time series data of two different time intervals; data, and convert multivariate time-series data into supervised data; train hand, foot and mouth disease prediction model based on supervised data; input real-time case data of hand, foot and mouth disease into the trained hand, foot and mouth disease prediction model to obtain real-time predicted hand, foot and mouth disease The number of cases of mouth disease. In this program, the time series data of different granularity time intervals of historical case data are counted, combined with finer-grained time series data, the accuracy of predicting the number of hand, foot and mouth disease cases based on equal time granularity is improved, and no additional data assistance is required, which improves the prediction efficiency. .

Description

Hand, foot and mouth disease prediction method, electronic equipment and medium using fine-grained data

technical field

The present application relates to the technical field of public health forecasting, in particular to a hand, foot and mouth disease forecasting method and device using fine-grained data, an electronic device, and a computer-readable medium.

Background technique

With the acceleration of the process of global economic integration, the increase of economic and communication activities, the increasingly frequent flow of people provides a favorable environment for the spread and outbreak of diseases, and public health problems are becoming more and more serious. At the same time, social and natural environments are also changing, and the increase in environmental pollution, natural disasters and other incidents that affect public health also increases the possibility of outbreaks of public health emergencies.

Hand, foot and mouth disease is one of the infectious diseases with the highest incidence rate, affecting public health security worldwide. Early prediction plays an important role in early warning and decision support for the prevention and control of infectious diseases. In addition, the control of hand, foot and mouth disease is also a concern of the government, medical institutions and ordinary people.

When traditional methods predict the number of HFMD cases in a certain time interval in the future, they are all based on observing the number of cases at the same time interval. The number of people, but the predictions of the above traditional methods are often unsatisfactory.

Contents of the invention

The purpose of this application is to provide a method and device for HFMD prediction using fine-grained data, an electronic device, and a computer-readable medium.

The first aspect of the present application provides a method for predicting hand, foot and mouth disease using fine-grained data, including:

S1. Obtain historical case data of hand, foot and mouth disease;

S2. Preprocessing the historical case data and making statistics into time series data of two different time intervals;

S3. Obtain multivariate time-series data according to time-aggregated time-series data of two different time intervals, and convert the multivariate time-series data into supervised data;

S4. According to the supervised data, train the hand, foot and mouth disease prediction model;

S5. Inputting the case data of HFMD collected in real time into the trained HFMD prediction model to obtain the number of cases of HFMD predicted in real time.

The second aspect of the present application provides a hand, foot and mouth disease prediction device using fine-grained data, including:

Obtain module, be used for obtaining the historical case data of hand, foot and mouth disease;

A preprocessing module, configured to preprocess the historical case data and make statistics into time series data of two different time intervals;

The aggregation module is used to aggregate the time-series data of the two different time intervals according to time to obtain multivariate time-series data, and convert the multivariate time-series data into supervised data;

Model training module, used for training hand, foot and mouth disease prediction model according to the supervised data;

The prediction module is used to input the case data of hand, foot and mouth disease collected in real time into the trained hand, foot and mouth disease prediction model to obtain the number of cases of hand, foot and mouth disease predicted in real time.

The third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor runs the computer program, it executes to realize The method described in the first aspect of the present application.

The fourth aspect of the present application provides a computer-readable medium on which computer-readable instructions are stored, and the computer-readable instructions can be executed by a processor to implement the method described in the first aspect of the present application.

Compared with the prior art, the hand, foot and mouth disease prediction method, device, electronic equipment and medium provided by this application can obtain historical case data of hand, foot and mouth disease using fine-grained data; preprocess the historical case data, and make statistics It is the time series data of two different time intervals; according to the time series data of the two different time intervals in advance, the multivariate time series data is obtained, and the multivariate time series data is converted into supervised data; according to the supervised data, Training the HFMD prediction model; inputting the HFMD case data collected in real time into the trained HFMD prediction model to obtain the real-time predicted incidence of HFMD. In this program, the time series data of different granularity time intervals of historical case data are counted to train the prediction model. Through the prediction model, combined with the more fine-grained time series data, the accuracy of predicting the number of cases of hand, foot and mouth disease based on equal time granularity is improved, and no additional With data assistance, the prediction efficiency is improved.

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the application. Also throughout the drawings, the same reference numerals are used to designate the same components. In the attached picture:

Fig. 1 shows a flow chart of a hand, foot and mouth disease prediction method using fine-grained data provided by some embodiments of the present application;

Fig. 2 shows a flow chart of a specific hand, foot and mouth disease prediction method using fine-grained data provided by some embodiments of the present application;

Fig. 3 shows a schematic diagram of a hand, foot and mouth disease prediction device using fine-grained data provided by some embodiments of the present application;

Fig. 4 shows a schematic diagram of an electronic device provided by some embodiments of the present application.

Detailed ways

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

It should be noted that, unless otherwise specified, technical terms or scientific terms used in this application shall have the usual meanings understood by those skilled in the art to which this application belongs.

In addition, the terms "first" and "second", etc. are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "include" and "have", as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but optionally also includes unlisted steps or units, or optionally further includes For other steps or units inherent in these processes, methods, products or apparatuses.

Due to the incubation period of the disease, most of the current HFMD prediction models cannot accurately predict the outbreak period of HFMD, so it is possible to mine the potential number of outbreaks from fine-grained data.

Therefore, this application collects the historical incidence cases of hand, foot and mouth disease, sets equal time intervals to the prediction target and a finer-grained time interval than the prediction target time interval, counts and obtains the time series of the number of cases in units of two time granularities , aggregate two time series with different time intervals, transform the time series data into supervised data, use the supervised data to train the time series neural network that fuses the two data, and finally use the trained model to provide an equal interval than only using and predicting the target more accurate forecasting of the time series.

Specifically, the embodiments of the present application provide a method and device for predicting hand, foot and mouth disease using fine-grained data, an electronic device, and a computer-readable medium, which will be described below with reference to the accompanying drawings.

Please refer to FIG. 1 , which shows a flow chart of a hand, foot and mouth disease prediction method using fine-grained data provided by some embodiments of the present application. As shown in the figure, the hand, foot and mouth disease prediction using fine-grained data method may include the following steps:

Step S101: Obtain historical case data of HFMD;

Step S102: Preprocessing the historical case data and making statistics into time series data of two different time intervals;

Step S103: Aggregating time-series data of the two different time intervals according to time to obtain multivariate time-series data, and converting the multivariate time-series data into supervised data;

Step S104: According to the supervised data, train the hand, foot and mouth disease prediction model;

Step S105: Input the case data of HFMD collected in real time into the trained HFMD prediction model to obtain the number of HFMD cases predicted in real time.

Specifically, in step S101, the historical case data of HFMD may include the number of cases and the time of cases. For example, the number of cases corresponding to each month, or the number of cases corresponding to each quarter, or the number of cases corresponding to each year.

In step S102, preliminary data cleaning is performed on historical case data, zero values are filtered, and invalid samples are removed. Normalize the cleaned data, compress the value of historical case data to a certain interval, and obtain the normalized data. Then the normalized data is counted as time series data of two different time intervals. Here, the time series data of two different time intervals may also be counted first, and then normalized, which is not limited in this application.

Among them, two different time intervals can be: the first time interval is the same as the target time interval, and the time series data counted according to the first time interval is called equal granularity time series data; the second time interval is smaller than the target time interval, and the The time series data counted according to the second time interval is called fine-grained time series data;

For example, if the target time interval is one year, the first time interval is one year, the second time interval can be set to one month, and the second time interval is a finer-grained setting of the first time interval.

Set the number of time intervals of the first type to M, then you can use [y ₁ , y ₂ , ..., y _M ] to represent time series data of equal granularity, and the number of patients in the upcoming time interval is expressed as

满足

其中x_t，i表示等粒度时刻t内的第i个细粒度数据。Set the first time interval to include N second time intervals, then [x ₁ , x ₂ ,…, x _M ] can be used to represent equal granularity time series data, where any

satisfy

Where x _{t, i} represent the i-th fine-grained data within the equal granularity time t.

Due to the significant difference in the range of data at different times, the present invention uses Min-Max to normalize the data to [0, 1]. The normalized formula is as follows:

表示某一粒度的时序数据，d′表示归一化后的时序数据，min(·)和max(·)分别表示输入向量的最小值和最大值。in,

Represents the time series data of a certain granularity, d' represents the normalized time series data, min(·) and max(·) represent the minimum and maximum values of the input vector, respectively.

In step S103:

In order to capture temporal properties from time series, the present invention transforms time series data into supervised data. Given a time series data, in the transformation process, the first several variables are used as the input variables of the model, and the latter variable is used as the output of the model.

Given the granular time series data [y ₁ , y ₂ ,...,y _M ] such as the number of cases, set the constant variable T as the number of time intervals that affect the number of cases in the next time interval. The time series is transformed into supervised data as:

Given the time series data [x _{1, i} , x _{2, i} , ..., x _{M, i} ] of the i-th fine-grained moment, set the constant variable T as the number of time intervals that affect the number of cases in the next time interval. The time series is transformed into supervised data as:

where i∈{1,...,N}.

According to the time, the time series data of two different time intervals are aggregated to obtain multivariate time series data, and the multivariate time series data is converted into supervised data, expressed as:

In step S104, train the hand, foot and mouth disease prediction model based on the above supervised data, please refer to Figure 2, which shows the flow of a specific hand, foot and mouth disease prediction method using fine-grained data provided by some embodiments of the present application picture:

As shown in the figure, the hand, foot and mouth disease prediction model includes: input layer, equal-grained sequence processing unit, fine-grained sequence processing unit, merge layer, fully connected layer and output layer;

Wherein, the input layer is used to input the supervised data, the equal granularity sequence processing unit is used to process the equal granularity time series data, the fine-grained sequence processing unit is used to process the fine-grained time series data, and then the combination layer and The fully connected layer correlates the data output by the equal-grained sequence processing unit and the fine-grained sequence processing unit to generate an initial prediction result, and the output layer denormalizes the initial prediction result to generate a final output prediction result.

The specific process is as follows:

I) The representation components of the equal granularity sequence processing unit include: the first GRU layer and the first thread layer. The first GRU layer is used to process the input and other granular time series data and output its hidden state. The first thread layer merges the hidden states and outputs the predicted values of the equal-grained sequence processing units.

The conversion of GRU is represented by f ₁ , and its formula is as follows:

h _t = f ₁ (h _t-1 , [y ₁ , y ₂ , . . . , y _T ]),

是隐藏状态，n表示隐藏状态的个数；h_t-1表示上一个时刻隐藏状态。利用等粒度时序数据产生预测值，可通过线程层进行转化，公式如下：in,

is the hidden state, n represents the number of hidden states; h _t-1 represents the hidden state at the previous moment. Use equal-grained time series data to generate forecast values, which can be converted through the thread layer. The formula is as follows:

c _y =w _y h _t +b _y ,

是等粒度数据侧的预测值，

为权重参数，

为偏置项。in

is the predicted value of the equal granularity data side,

is the weight parameter,

is a bias term.

II) The representation components of the equal granularity sequence processing unit include: the second thread layer, the softmax layer, the second GRU and the third thread layer. The second thread layer linearly converts the input fine-grained data, the softmax layer is used to strengthen the fine-grained data input at important moments, the second GRU layer is used to describe the timing characteristics, and the third thread layer is used to merge hidden states and output fine-grained components predicted value of .

Highlighting periodic events by linear weighting is a key method to enhance the extraction of input information. The linear weights of the input features can be summarized as follows:

u=∑w _u *[x ₁ , x ₂ , . . . , x _T ]+b _u ,

表示输入的T等粒度时间间隔内对应的细粒度数据,

是权重，b_u是偏差项，

为线性化输出。in,

Indicates the fine-grained data corresponding to the input T and other granular time intervals,

is the weight, b _u is the bias term,

is the linearized output.

After linear weighting, there are data that are weakly or uncorrelated with the predicted target time. In order to ensure that the model can extract the potential information of all input elements, softmax is used to weaken the influence of weakly correlated or irrelevant data and increase the weight of relevant data. The formula is:

p=softmax(e),

表示该层输出。对于p中的任一元素有：in,

Indicates the output of this layer. For any element in p there are:

Among them, exp(·) represents the exponential function, and e _{j, t} represent the elements in e.

The second GRU layer is used to dynamically extract the timing features in p, and its formula is as follows:

h′ _t = f ₂ (h′ _t-1 , p)

是隐藏状态；h′_t-1表示上一个时刻隐藏状态。利用细粒度时序数据产生预测值，可通过线程层进行转化，公式如下：in,

is the hidden state; h′ _t-1 represents the hidden state at the previous moment. Using fine-grained time series data to generate forecast values can be converted through the thread layer. The formula is as follows:

c _e =w _e h′ _t +b _e ,

是细粒度数据侧的预测值，

为权重，b_e是偏置项。in

is the predicted value on the fine-grained data side,

is the weight, and b _e is the bias term.

Ⅲ) In the merge layer, the output of the two processing units (equal-grained sequence processing unit and fine-grained sequence processing unit) represents data merging, and then through a fully connected layer, the equal-grained time-series data and fine-grained time-series data are combined to correlate double The output of the side (two processing units) combines fine-grained and equal-grained prediction results through the output layer. This step can be summarized as follows:

是双侧输出的合并向量，

是这些输出的权重，

是预测的下一个时间间隔的门诊病人数，

是偏置项。in

is the combined vector of two-sided outputs,

are the weights of these outputs,

is the predicted number of outpatients in the next time interval,

is a bias item.

The denormalization operation is performed on the output prediction results, and the calculation formula is as follows:

d=d'*(max(d)-min(d))+min(d),

The denormalization formula is applied in the method to correct the final forecast results produced by the forecast model.

V) The above hand, foot and mouth disease prediction model uses Mean squared error (MSE) as the loss function of the prediction model, that is, the objective function. Through iterative training, the loss of the objective function is minimized, and an optimal prediction model is established.

Specifically, set the objective function as MSE, set the time window size as T and the GRU hidden state parameter n. According to the data generated in step S103 and step S104 and the corresponding prediction model, the prediction model is trained. Using the denormalization function, restore the output of the prediction model to the predicted number of cases. Adjust the model parameters T and n according to the output results to obtain the optimal model.

In step S105, input the case data of HFMD collected in real time into the trained HFMD prediction model to obtain the number of HFMD cases predicted in real time.

As shown in Figure 2, after denormalizing the output of the prediction model, the prediction results are displayed to public health personnel for disease prevention.

The above-mentioned hand, foot and mouth disease prediction method using fine-grained data can be applied to the client. In the embodiment of the present application, the client may include hardware or software. When the client includes hardware, it may be various electronic devices with display screens and supporting information interaction, for example, may include but not limited to smart phones, tablet computers, laptop computers, desktop computers and the like. When the client includes software, it can be installed in the above-mentioned electronic device, and it can be implemented as multiple software or software modules, or as a single software or software module. No specific limitation is made here.

Compared with the prior art, the HFMD prediction method using fine-grained data provided by the present invention has the following advantages:

1. No need for external data sources

The present invention counts the historical morbidity data of hand, foot and mouth disease by setting an equal time interval to the prediction target and a finer-grained time interval than the time interval of the prediction target, and uses finer-grained time series data to improve the accuracy of the prediction target based on equal time granularity. Spend. From the perspective of data, the present invention does not require additional data assistance; from the perspective of algorithm, the present invention captures fine-grained disease conditions to judge future trends.

2. Discover the crisis from the details

The data used in this method is only HFMD case data. The essence of the present invention is to count the number of infected people in a finer-grained time period, so as to predict the outbreak situation in a longer time period in the future, so as to see the big from the small. Due to the incubation period of the disease, it is possible to mine potential outbreak numbers from fine-grained data.

In the above embodiments, a method for predicting HFMD using fine-grained data is provided. Correspondingly, the present application also provides a device for predicting HFMD using fine-grained data. The HFMD prediction device using fine-grained data provided by the embodiment of the present application can implement the above-mentioned HFMD prediction method using fine-grained data. The HFMD prediction device using fine-grained data can use software, hardware, or a combination of hardware and software. way to achieve. For example, the device for predicting hand, foot and mouth disease using fine-grained data may include integrated or separate functional modules or units to perform corresponding steps in the above-mentioned methods. Please refer to FIG. 3 , which shows a schematic diagram of a hand, foot and mouth disease prediction device using fine-grained data provided by some embodiments of the present application. Since the device embodiment is basically similar to the method embodiment, the description is relatively simple, and for relevant parts, refer to the part of the description of the method embodiment. The device embodiments described below are illustrative only.

As shown in Figure 3, the HFMD prediction device 10 using fine-grained data may include:

Obtaining module 101, for obtaining the historical case data of hand, foot and mouth disease;

A preprocessing module 102, configured to preprocess the historical case data and make statistics into time series data of two different time intervals;

The aggregation module 103 is used to aggregate the time series data of the two different time intervals according to time to obtain multivariate time series data, and convert the multivariate time series data into supervised data;

Model training module 104, used for training hand, foot and mouth disease prediction model according to the supervised data;

The prediction module 105 is configured to input the case data of HFMD collected in real time into the trained HFMD prediction model to obtain the number of cases of HFMD predicted in real time.

In some implementations of the embodiments of the present application, the preprocessing module 102 is specifically configured to: perform data cleaning on the historical case data, and perform normalization processing on the cleaned data.

In some implementations of the embodiments of the present application, among the two different time intervals, the first time interval is the same as the target time interval, and the time series data counted according to the first time interval is called equal granularity time series data; the second time series If the time interval is smaller than the target time interval, the time series data counted according to the second time interval is called fine-grained time series data.

The device 10 for predicting HFMD using fine-grained data provided in the embodiment of the present application is based on the same inventive concept as the method for predicting HFMD using fine-grained data provided in the previous embodiments of the present application, and has the same beneficial effect.

The embodiment of the present application also provides an electronic device corresponding to the hand, foot and mouth disease prediction method using fine-grained data provided in the foregoing embodiment, and the electronic device may be an electronic device for a client, such as a mobile phone, a notebook computer, a tablet Computers, desktop computers, etc., to execute the above-mentioned hand, foot and mouth disease prediction method using fine-grained data.

Please refer to FIG. 4 , which shows a schematic diagram of an electronic device provided by some embodiments of the present application. As shown in Figure 4, described electronic device 20 comprises: processor 200, memory 201, bus 202 and communication interface 203, described processor 200, communication interface 203 and memory 201 are connected by bus 202; Stored in the memory 201 There is a computer program that can run on the processor 200, and when the processor 200 runs the computer program, it executes the hand, foot and mouth disease prediction method using fine-grained data provided by any one of the above-mentioned embodiments of the present application.

Wherein, the memory 201 may include a high-speed random access memory (RAM: Random Access Memory), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 203 (which may be wired or wireless), and Internet, wide area network, local network, metropolitan area network, etc. can be used.

The bus 202 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus can be divided into address bus, data bus, control bus and so on. Wherein, the memory 201 is used to store a program, and the processor 200 executes the program after receiving an execution instruction, and the method for predicting hand, foot and mouth disease using fine-grained data disclosed in any one of the above-mentioned embodiments of the present application can be Applied in the processor 200, or implemented by the processor 200.

The processor 200 may be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above method may be implemented by an integrated logic circuit of hardware in the processor 200 or instructions in the form of software. The above-mentioned processor 200 can be a general-purpose processor, including a central processing unit (Central Processing Unit, referred to as CPU), a network processor (Network Processor, referred to as NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. Various methods, steps, and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. The steps of the method disclosed in the embodiments of the present application may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register. The storage medium is located in the memory 201, and the processor 200 reads the information in the memory 201, and completes the steps of the above method in combination with its hardware.

The electronic device provided in the embodiment of the present application is based on the same inventive concept as the method for predicting hand, foot and mouth disease using fine-grained data provided in the embodiment of the present application, and has the same beneficial effect as the method adopted, operated or implemented.

The embodiments of the present application also provide a computer-readable medium corresponding to the method for predicting hand, foot and mouth disease using fine-grained data provided in the foregoing embodiments, on which a computer program (ie, a program product) is stored, and the computer program is When the processor is running, it will execute the hand, foot and mouth disease prediction method using fine-grained data provided by any of the above implementations.

It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random Access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other optical and magnetic storage media will not be repeated here.

The computer-readable storage medium provided by the above-mentioned embodiments of the present application is based on the same inventive concept as the hand-foot-and-mouth disease prediction method using fine-grained data provided by the embodiments of the present application, and has the functions adopted, run or realized by the application program stored therein. The same beneficial effect of the method.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit it; although the application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present application. All of them should be covered by the scope of the claims and description of the present application.

Claims (6)

1. A hand-foot-and-mouth disease prediction method using fine-grained data is characterized by comprising the following steps:

s1, acquiring historical case data of hand-foot-and-mouth disease;

s2, preprocessing the historical case data, and counting the historical case data into two time sequence data with different time intervals;

s3, performing data aggregation on the time sequence data of the two different time intervals according to time to obtain multivariate time sequence data, and converting the multivariate time sequence data into supervised data;

s4, training a hand-foot-and-mouth disease prediction model according to the supervised data;

s5, inputting the real-time collected case data of the hand-foot-and-mouth disease into the trained hand-foot-and-mouth disease prediction model to obtain the number of the real-time predicted patients with the hand-foot-and-mouth disease;

in the two different time intervals in the step S2, the first time interval is the same as the target time interval, and the time series data counted according to the first time interval is called as equal-granularity time series data; the second time interval is smaller than the target time interval, and the time sequence data counted according to the second time interval is called fine-grained time sequence data;

setting the number of the first time intervals to be M, using [ y ₁ ,y ₂ ,…,y _M ]Representing equal-granularity time series data;

setting the first time interval to contain N second time intervals, using [ x ] ₁ ,x ₂ ,…,x _M ]The fine-grained data corresponding to the input M equal-grained time interval is represented arbitrarily

Satisfy the requirement of

Wherein x _t,i Indicating equal granularity within time tThe ith fine-grained data of (1);

the preprocessing step in step S2 includes:

carrying out data cleaning on the historical case data, and carrying out normalization processing on the cleaned data;

the step S3 includes:

setting a constant variable T as the number of time intervals influencing the number of the persons who suffer from the diseases in the next time interval, converting the equal-granularity time series data, and expressing as follows:

setting time series data [ x ] of ith fine-grained moment _1,i ,x _2,i ,…,x _M,i ]After the time series data is converted, the time series data is expressed as:

wherein, i belongs to {1, …, N };

carrying out data aggregation on the time sequence data of two different time intervals according to time to obtain multivariate time sequence data, and converting the multivariate time sequence data into supervised data, wherein the expression is as follows:

the specific treatment process is as follows:

i) the presentation component of the equal-granularity sequence processing unit comprises: a first GRU layer and a first thread layer; the first GRU layer is used for processing input equal-granularity time sequence data and outputting a hidden state of the input equal-granularity time sequence data; the first thread layer merges the hidden states and outputs the predicted value of the equal-granularity sequence processing unit;

conversion of GRU ₁ Expressed, its formula is as follows:

h _t ＝f ₁ (h _t-1 ,[y ₁ ,y ₂ ,…,y _T ])，

wherein,

is a hidden state, n represents the number of hidden states; h is _t-1 Representing a previous time hidden state; the equal-granularity time sequence data is used for generating a predicted value, and the predicted value can be converted through a thread layer, wherein the formula is as follows:

c _y ＝w _y h _t +b _y ，

wherein

Is a predicted value of the equal-granularity data side,

in order to be a weight parameter, the weight parameter,

is a bias term;

II) the representation part of the equal-granularity sequence processing unit comprises: the system comprises a second thread layer, a softmax layer, a second GRU and a third thread layer; the second thread layer carries out linear conversion on input fine-grained data, the softmax layer is used for strengthening fine-grained data input at important moments, the second GRU layer is used for describing time sequence characteristics, and the third thread layer is used for merging hidden states and outputting predicted values of fine-grained components;

highlighting periodic events by linear weighting is a key method for enhancing input information extraction; the linear weights of the input elements can be summarized as follows:

u＝∑w _u *[x ₁ ,x ₂ ,…,x _T ]+b _u ，

wherein,

representing the corresponding fine-grained data in the input T equal-grained time interval,

is a weight, b _u Is the term of the deviation in the sense that,

is a linearized output;

after linear weighting, the softmax layer is utilized to weaken the influence of weakly correlated or uncorrelated data and improve the weight of correlated data, and the formula is as follows:

p＝softmax(e)，

wherein,

representing softmax layer output; for any element in p are:

wherein exp (·) represents an exponential function;

the second GRU layer is used to dynamically extract the timing characteristics in p, which is formulated as follows:

h' _t ＝f ₂ (h' _t-1 ,p)

wherein,

is a hidden state; h' _t-1 Representing a previous time hidden state; the fine-grained time series data are used for generating a predicted value, and conversion can be carried out through a thread layer, and the formula is as follows:

c _e ＝w _e h' _t +b _e ，

wherein

Is a predicted value of the fine-grained data side,

is a weight, b _e Is a bias term;

III) merging the output representation data of the medium-granularity sequence processing unit and the fine-granularity sequence processing unit in the merging layer, then combining the equal-granularity time sequence data and the fine-granularity time sequence data through a full connection layer to associate the output of the two processing units, and outputting a prediction result combining the fine granularity and the equal granularity through an output layer; this step can be summarized as follows:

wherein

Is a merged vector that is output from both sides,

it is the weight of these outputs that is,

is the predicted number of outpatients for the next time interval,

is a bias term;

performing inverse normalization operation on the output prediction result, wherein the calculation formula is as follows:

d＝d'*(max(d)-min(d))+min(d)，

the inverse normalization formula is applied in the method to correct the final prediction result generated by the prediction model.

2. The method of claim 1, wherein the hand-foot-and-mouth disease prediction model comprises: the system comprises an input layer, an equal-granularity sequence processing unit, a fine-granularity sequence processing unit, a merging layer, a full-connection layer and an output layer;

the input layer is used for inputting the supervised data, the equal-granularity sequence processing unit is used for processing equal-granularity time series data, the fine-granularity sequence processing unit is used for processing fine-granularity time series data, then a merging layer and a full-connection layer are used for mutually correlating the data output by the equal-granularity sequence processing unit and the data output by the fine-granularity sequence processing unit to generate an initial prediction result, and the output layer is used for carrying out reverse normalization on the initial prediction result to generate a final output prediction result.

3. The method according to claim 2, characterized by using the mean square error as a loss function of the hand-foot-and-mouth disease prediction model.

4. An apparatus for predicting hand-foot-and-mouth disease using fine-grained data, comprising:

the acquisition module is used for acquiring historical case data of the hand-foot-and-mouth disease;

the preprocessing module is used for preprocessing the historical case data and counting the historical case data into two time sequence data with different time intervals;

the aggregation module is used for carrying out data aggregation on the time sequence data of the two different time intervals according to time to obtain multivariate time sequence data and converting the multivariate time sequence data into supervised data;

the model training module is used for training a hand-foot-and-mouth disease prediction model according to the supervised data;

the prediction module is used for inputting the real-time collected case data of the hand-foot-and-mouth disease into the trained hand-foot-and-mouth disease prediction model to obtain the number of the real-time predicted patients of the hand-foot-and-mouth disease;

the preprocessing module is also used for calling the time sequence data counted according to the first time interval as equal-granularity time sequence data in two different time intervals, wherein the first time interval is the same as the target time interval; the second time interval is smaller than the target time interval, and the time sequence data counted according to the second time interval is called fine-grained time sequence data;

setting the number of the first time intervals to be M, using [ y ₁ ,y ₂ ,…,y _M ]Indicating equal grainDegree time sequence data;

setting the first time interval to contain N second time intervals, using [ x ] ₁ ,x ₂ ,…,x _M ]The fine-grained data corresponding to the input M equal-grained time interval is represented arbitrarily

Satisfy the requirement of

Wherein x _t,i Representing the ith fine-grained data within the equal-grained time t;

the historical case data processing device is used for cleaning the historical case data and carrying out normalization processing on the cleaned data;

and the aggregation module is also used for setting a constant variable T as the number of time intervals influencing the number of the people suffering from the diseases in the next time interval, and converting the equal-granularity time sequence data to be expressed as:

setting time series data [ x ] of ith fine-grained moment _1,i ,x _2,i ,…,x _M,i ]After the time series data is converted, the time series data is expressed as:

wherein, i belongs to {1, …, N };

carrying out data aggregation on the time sequence data of two different time intervals according to time to obtain multivariate time sequence data, and converting the multivariate time sequence data into supervised data, wherein the expression is as follows:

the specific treatment process is as follows:

i) the presentation component of the equal-granularity sequence processing unit comprises: a first GRU layer and a first thread layer; the first GRU layer is used for processing input equal-granularity time sequence data and outputting a hidden state of the input equal-granularity time sequence data; the first thread layer merges the hidden states and outputs the predicted value of the equal-granularity sequence processing unit;

conversion of GRU ₁ Expressed, its formula is as follows:

h _t ＝f ₁ (h _t-1 ,[y ₁ ,y ₂ ,…,y _T ])，

wherein,

is a hidden state, n represents the number of hidden states; h is _t-1 Representing a previous time hidden state; the equal-granularity time sequence data is used for generating a predicted value, and the predicted value can be converted through a thread layer, wherein the formula is as follows:

c _y ＝w _y h _t +b _y ，

wherein

Is a predicted value of the equal-granularity data side,

in order to be a weight parameter, the weight parameter,

is a bias term;

II) the representation part of the equal-granularity sequence processing unit comprises: the system comprises a second thread layer, a softmax layer, a second GRU and a third thread layer; the second thread layer carries out linear transformation on the input fine-grained data, the softmax layer is used for strengthening fine-grained data input at an important moment, the second GRU layer is used for describing time sequence characteristics, and the third thread layer is used for merging hidden states and outputting a predicted value of a fine-grained component;

highlighting periodic events by linear weighting is a key method for enhancing input information extraction; the linear weights of the input elements can be summarized as follows:

u＝∑w _u *[x ₁ ,x ₂ ,…,x _T ]+b _u ，

wherein,

representing the corresponding fine-grained data in the input T equal-grained time interval,

is a weight, b _u Is the term of the deviation in the sense that,

is a linearized output;

after linear weighting, the softmax layer is utilized to weaken the influence of weakly correlated or uncorrelated data and improve the weight of correlated data, and the formula is as follows:

p＝softmax(e)，

wherein,

representing softmax layer output; for any element in p are:

wherein exp (·) represents an exponential function;

the second GRU layer is used to dynamically extract the timing characteristics in p, which is formulated as follows:

h' _t ＝f ₂ (h' _t-1 ,p)

wherein,

is a hidden state; h' _t-1 Representing a previous time hidden state; the fine-grained time series data are used for generating a predicted value, and conversion can be carried out through a thread layer, and the formula is as follows:

c _e ＝w _e h' _t +b _e ，

wherein

Is a predicted value of the fine-grained data side,

is a weight, b _e Is a bias term;

III) merging the output representation data of the two processing units (the equal-granularity sequence processing unit and the fine-granularity sequence processing unit) in the merging layer, combining the equal-granularity time sequence data and the fine-granularity time sequence data through a full connection layer to correlate the output of the two sides (the two processing units), and outputting a prediction result combining the fine granularity and the equal granularity through an output layer; this step can be summarized as follows:

wherein

Is a merged vector that is output from both sides,

it is the weight of these outputs that is,

is the predicted number of outpatients for the next time interval,

is a bias term;

performing inverse normalization operation on the output prediction result, wherein the calculation formula is as follows:

d＝d'*(max(d)-min(d))+min(d)，

the anti-normalization formula is applied in the method to correct the final prediction result generated by the prediction model.

5. An electronic device, comprising: memory, processor and computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method according to any of claims 1 to 3.

6. A computer readable medium having computer readable instructions stored thereon which are executable by a processor to implement the method of any one of claims 1 to 3.

Images