TW201419161A - Load prediction method and electronic apparatus - Google Patents

Abstract

A load prediction method and an electronic device. In the load prediction method, a plurality of resource load records of an electronic device are recorded in a plurality of time periods, respectively. When the predicted time point is received, the time difference between the predicted time point and the current time point is calculated, wherein the predicted time point is greater than the current time point on the time axis. And, it is judged whether the time difference is smaller than the threshold value. If the time difference is less than the threshold, the regression analysis prediction procedure is performed based on the resource load record. If the time difference is not less than the threshold, the group analysis prediction process is performed according to the resource load record.

Description

Load prediction method and electronic device

The present invention relates to a resource management technique, and in particular to a load prediction method and an electronic device.

With the advancement of computer computing power, a lot of information is accumulated in life, such as user consumption behavior information, road condition information, and sensing data. Faced with such a large amount of data, the ability to use cloud computing to analyze large amounts of data plays an increasingly important role. Many cloud service providers that provide computing power have emerged, such as the well-known cloud service provider Amazon Elastic Compute Cloud (Amazon EC2). Amazon EC2 is a web service whose main function is to provide computing power of various sizes on the cloud computing platform, providing developers with a convenient computing environment to perform large-scale computing.

However, the system resource allocation method of the above cloud computing platform cannot actually meet the requirements of the user completely, for example, the system resources cannot be fully utilized. Therefore, in order to improve user satisfaction and the overall computing performance of the cloud computing platform, a dynamic resource management mechanism is usually constructed in the cloud computing platform to effectively allocate resources to all users.

The current resource management mechanism is achieved, for example, by means of load prediction. Common load prediction methods are, for example, a regression analysis method and a neural network-like method. The regression analysis method is based on historical data to find the polynomial closest to the data, but due to the general regression score The analysis method uses multi-dimensional data for analysis, and needs to record various data to increase the accuracy, so it requires a large storage space and system resources of the computer system. In addition, the Artificial Neural Network (ANN) will continuously correct the prediction model, and must predict the predicted value of the load at the next time point based on the corrected prediction model, so it is not suitable for long-term use. Load forecasting.

In view of the above, how to effectively allocate the resource allocation in the system has become one of the problems that manufacturers are eager to solve.

The present invention provides a load prediction method and an electronic device, which can effectively predict a resource load record of an electronic device and improve the working performance of the electronic device.

The invention provides a load prediction method suitable for an electronic device. In the load prediction method, a plurality of resource load records of an electronic device are recorded in a plurality of time periods, respectively. Receive the predicted time point. A time difference between the predicted time point and the current time point is calculated, wherein the predicted time point is greater than the current time point on the time axis. Determine if the time difference is less than the threshold. If the time difference is less than the threshold, a regression-based prediction procedure is performed according to the resource load record. If the time difference is not less than the threshold value, a clustering-based prediction procedure is performed according to the resource load record.

In an embodiment of the present invention, the step of performing a regression analysis prediction procedure according to the resource load record includes, for the recorded location The resource load record performs a regression analysis algorithm to obtain a prediction model, and the predicted value corresponding to the predicted time point is taken out in the prediction model.

In an embodiment of the present invention, the step of performing a group analysis prediction program according to the resource load record includes dividing each of the time periods into a plurality of time segments, so that each of the resource load records is divided. Multiple pieces of data. One of the predicted time segments corresponding to the predicted time point is selected in the time segment. The data segments of the predicted time segment during the time period are clustered to group the data segments into a plurality of clusters. From the cluster, select one of the clusters with the largest number of pens. The average of the data segments included in the selected cluster is calculated as a predicted value for the predicted time point.

In an embodiment of the present invention, the step of performing cluster analysis on the data segments in the time period for each of the time segments includes calculating each data in each of the time segments A similarity between the segments and cluster analysis based on similarity.

In an embodiment of the present invention, the step of recording the resource load record of the electronic device in the time period, respectively, includes obtaining a resource usage amount according to a sampling rate in each of the time periods. Therefore, each resource load record includes multiple resource usage.

The invention provides an electronic device, which comprises a recording module, a receiving module, a time calculating module, a determining module, a regression analysis module and a group analysis module. The recording module records a plurality of resource load records of the electronic device in a plurality of time periods. The receiving module receives the predicted time point. The time calculation module is used to calculate the time difference between the predicted time point and the current time point, wherein The predicted time point is greater than the current time point on the time axis. The determining module is configured to determine whether the time difference is less than a threshold value. The regression analysis module is configured to perform a regression analysis prediction process based on the resource load record. The group analysis module is configured to perform a group analysis prediction process according to the resource load record. Wherein, if the determination time difference is less than the threshold value, the determining module notifies the regression analysis module to execute the regression analysis prediction program; if the determination time difference is not less than the threshold value, the determination module notifies the group analysis module to execute the group analysis prediction program.

In an embodiment of the present invention, the regression analysis module performs a regression analysis algorithm on the recorded resource load record to obtain a prediction model, so as to extract a prediction value corresponding to the prediction time point in the prediction model.

In an embodiment of the present invention, the group analysis module further includes a division module, a prediction module, a grouping module, a selection module, and a prediction calculation module. The partitioning module divides each of the time periods into a plurality of time segments such that each of the resource load records is divided into a plurality of data segments. The prediction module selects one of the predicted time segments corresponding to the predicted time point in the time segment. The clustering module is coupled to the partitioning module to perform a cluster analysis on the data segments of the predicted time segment in the time period to group the data segments into a plurality of clusters. Select the module from the cluster and select one of the clusters with the largest number of segments. The predictive computing module calculates an average of the pieces of data included in the selected cluster as a predicted value for the predicted time point.

In an embodiment of the present invention, the grouping module calculates the similarity between each piece of data in each of the time segments, and according to the similarity Perform a cluster analysis.

In an embodiment of the invention, the recording module obtains a resource usage amount according to a sampling rate in each of the time periods.

Based on the above, the present invention proposes a hybrid prediction method, combined with regression analysis prediction and group analysis prediction, so that the two methods complement each other's deficiencies. Regression analysis is used for short-term predictions, and group analysis predictions are used for long-term predictions to improve prediction accuracy.

The above described features and advantages of the present invention will be more apparent from the following description.

The present invention provides a load prediction method and an electronic device that perform different prediction procedures according to a time difference between a predicted time point and a current time point, thereby effectively predicting a resource load record of the electronic device at a predicted time point. In order to make the content of the present invention easier to understand, the following examples are given as examples in which the present invention can be implemented.

First embodiment

1 is a flow chart of a load prediction method according to a first embodiment of the present invention. The load prediction method in this embodiment is applicable to an electronic device, wherein the electronic device has a processing unit, and the processing unit can execute a plurality of code segments stored in the storage unit, thereby implementing the steps of the following load prediction method.

Referring to FIG. 1, in step S102, the processing unit is in multiple times. The inter-cycle records a plurality of resource load records of the electronic device. Specifically, for example, the processing unit takes a sampling rate every d minutes (d is a positive integer) to separately capture a resource usage amount at a plurality of time points. In other words, each resource load record includes multiple resource usage. For example, assuming a time period of 24 hours, and assuming that the processing unit fetches each resource load record every 2 minutes, each resource load record includes 720 pens of resource usage.

In other words, the processing unit obtains multiple resource usages in each time period according to the sampling rate and records them as resource load records. That is to say, each time period has a corresponding resource load record. Assuming a time period of one day, a resource load record, that is, the history of the day, is recorded every day.

Thereafter, in step S104, the processing unit receives the predicted time point. Here, the predicted time point is greater than the current time point on the time axis, that is, the predicted time point is a point in time that occurs in the future. And, in step S106, the processing unit calculates a time difference between the predicted time point and the current time point. Specifically, assuming that the predicted time point is 9:00 am on May 19, and the current time point is 7:00 am on May 19, the time difference is 2 hours.

Next, in step S108, the processing unit determines whether the time difference is less than a threshold value. The processing unit can set the threshold value according to the user's setting. If the time difference between the predicted time point and the current time point is less than the threshold value, step S110 is performed; if the time difference between the predicted time point and the current time point is not less than the threshold value, step S112 is performed.

In step S110, the processing unit records according to the resource load Perform a regression analysis prediction program. In detail, in the regression analysis prediction program, the processing unit performs a regression analysis algorithm using the recorded resource load record to obtain a prediction model, and the processing unit can use the prediction model to predict the electronic device. Resource load record at the predicted time point. For example, a statistical regression analysis algorithm is used to obtain a prediction module, that is, to find a function (regression estimation formula) that best represents all resource usage, and this function represents between time and resource usage. relationship. Therefore, the predicted time point and the current time point are brought into the prediction model (ie, the regression estimation formula), thereby obtaining the predicted value of the predicted time point, that is, the predicted resource usage.

If the time difference is not less than the threshold, then as shown in step S112, the processing unit performs a group analysis prediction procedure according to the resource load record. Specifically, in the group analysis prediction program, the processing unit performs cluster analysis on the resource load record of the predicted time segment corresponding to the predicted time point to divide the resource load record in the predicted time segment into Multiple clusters, and in the cluster, the average of one of the clusters with the largest number of segments is selected as the predicted value of the predicted time point.

Based on the above, the embodiment can use the plurality of resource load records recorded by the electronic device in a plurality of time periods to predict the resource usage of the electronic device at the predicted time point. In this embodiment, based on the time difference between the predicted time point and the current time point, one of the regression analysis prediction program and the group analysis prediction program is selected to predict the resource usage of the electronic device at the predicted time point. Wherein, when the predicted time point is close to the current time point (ie, the time difference is less than the threshold value), the regression analysis predictive program can be used to predict the capital. Source usage. Conversely, when the predicted time point is not close to the current time point (ie, the time difference is not less than the threshold), the resource usage can be predicted by the group analysis prediction program. Thereby, according to the recorded resource load record, the resource usage in different application requirements and the operating environment can be predicted, thereby effectively predicting the predicted value of the resource usage occurring at a future time point, according to which the predicted value is based To adjust the resource usage allocation in the electronic device. Therefore, the electronic device can avoid the situation of uneven load and can have higher work efficiency.

An embodiment will be described below to explain the above-mentioned group analysis prediction program in detail.

Second embodiment

2 is a flow chart of a group analysis prediction program according to a second embodiment of the present invention. In this embodiment, the electronic device stores a plurality of resource load records recorded in a plurality of time periods, that is, each time period has a corresponding one of the resource load records.

Referring to FIG. 2, in step S202, the processing unit divides each time period into a plurality of time segments, so that each resource load record is divided into a plurality of data segments. For example, if the time period is one day, and the processing unit divides the day into 4 time segments, the time of each time segment is 6 hours, that is, the resource load record of each day is divided into 0:00-6. Six pieces of data such as :00, 6:00-12:00, 12:00-18:00, and 18:00-24:00. Taking the sampling rate as 2 minutes as an example, each data segment includes 180 resource usages. Also, in each time zone The segment will include a plurality of data segments of the time period in the time segment. In terms of 10 resource load records, that is, 10 days of history, 10 pieces of data will be included in the time zone 6:00-12:00, that is, the data of 6:00-12:00 in these 10 days. Fragment. The rest is also like this.

Next, in step S204, the processing unit selects one of the predicted time segments corresponding to the predicted time point in the time segment. That is, the processing unit selects the time segment including this predicted time point as the predicted time segment.

Then, in step S206, the processing unit performs cluster analysis on the data segments of the predicted time segment in the above time period to group the data segments into a plurality of clusters. In detail, when performing the cluster analysis, the processing unit may calculate the sum of the distances between the resource load records acquired at the same time point in each data segment in each time segment, and calculate the total distance using the distance. A similarity between each piece of data in each of the time segments, and a piece of data having a greater degree of similarity is divided into the same cluster, wherein the processing unit is like an inter-sequence editing distance (Edit Distance on Real) Sequence, EDR) algorithm, Longest Common Subsequence (LCS) algorithm, Edit distance with Real Penalty (ERP) algorithm or Dynamic Time Warping (DTW) algorithm, etc. The present invention is not limited to the above, but it is preferable that the processing unit of the embodiment calculates the similarity through the EDR algorithm.

In addition, the above cluster analysis is, for example, a k-means clustering method and a spatial clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBScan), or Hierarchical clustering, to divide the data segments in each time segment of each time period into multiple clusters. This embodiment is preferably a k-means clustering method, but this embodiment does not limit the method of cluster analysis.

Next, in step S208, the processing unit selects one of the clusters with the largest number of segments from the cluster divided by the processing unit. Moreover, in step S210, the processing unit calculates an average value of the data segments included in the selected cluster as a predicted value of the predicted time point. That is, the processing unit averages the resource load records in the data segments included in the selected cluster, and finds the average of the clusters as the resource load record at the predicted time point.

For example, with 30 resource load records, the time period is 1 day, divided into 6 time segments each day (0:00-4:00, 4:00-8:00, 8:00-12:00) For example, 12:00-16:00, 16:00-20:00, 20:00-24:00), assuming that the predicted time point is 3:00, the selected predicted time zone is 0:00. -4:00. According to this, in the 30 resource load records, the data segments (30 pens) of the predicted time segment 0:00-4:00 are taken out, and 30 pieces of data segments are clustered. Assume that cluster A, cluster B, and cluster C are obtained, which include 10 segments, 15 pens, and 5 segments of data segments, respectively. According to this, the cluster C corresponding to 15 pieces of data is taken out. After that, the average of 15 data segments of cluster C is calculated. Alternatively, only the average value of the cluster C at the predicted time point 3:00 is calculated as the predicted value. Based on the above, the present embodiment can utilize the electronic device for a plurality of time periods. A plurality of resource load records are recorded to predict a resource load record of the electronic device at the predicted time point. In particular, when the predicted time point is not close to the current time point (ie, the time difference is not less than the threshold value), the resource load record may be predicted by the group analysis prediction program, thereby selecting the predicted time segment corresponding to the predicted time point. The average of most of the recorded resource load records is used as a predictor of the resource load record at the predicted time point. According to this, the predicted value of the resource load record occurring at a future time point can be effectively predicted, and the resource use allocation in the electronic device can be adjusted according to the predicted value, thereby achieving the purpose of improving the work efficiency.

For the above load prediction method, the present invention also provides a corresponding electronic device, so that the method can be applied to a personal device such as a personal computer, a notebook computer, a tablet computer, a personal digital assistant (PDA), a server, a mobile phone, and the like. In the following, an embodiment will be described in detail.

Third embodiment

3 is a block diagram of an electronic device in accordance with a third embodiment of the present invention. Referring to FIG. 3, the electronic device 300 is, for example, a personal computer, a notebook computer, a tablet computer, a personal digital assistant (PDA), a server, a mobile phone, or the like. The present invention does not impose the type of the electronic device 300. limit.

Referring to FIG. 3 , the electronic device 300 includes a recording module 302 , a receiving module 304 , a time computing module 306 , a determining module 308 , a regression analysis module 310 , and a group analysis module 312 . Functional description of each module above as follows.

The recording module 302 is configured to record a plurality of resource load records, that is, history records, of the electronic device 300. The resource load record is, for example, a central processing unit (CPU) usage rate, a memory usage rate, a page file (PF) usage amount, or a network usage amount, and is not limited to the above. In addition, the recording module 302 can record the resource load of the electronic device 300 according to the recording time set by the user, wherein the recording time is, for example, one week or several weeks, one or more months, and the like. The example does not limit the length of the recording time. That is to say, in the above recording time, the recording module 302 can obtain different resource load records according to the resource usage of the electronic device 300 in different application requirements and the operating environment.

For example, when the electronic device 300 enters the Sleep Configuration mode or the Hibernate mode of the Advanced Configuration and Power Interface (ACPI), the electronic device 300 is maintained at a lower performance. The work is performed so that the recording module 302 can record lower resource usage. On the other hand, when the electronic device 300 enters the normal working mode, for example, when the electronic device 300 executes software, an application program, or other executing programs, since the electronic device 300 maintains a high-performance working operation, the recording module 302 Higher resource usage can be recorded.

The receiving module 304 is configured to receive a predicted time point, wherein the predicted time point is greater than the current time point on the time axis. That is to say, the present embodiment can predict the resource usage of the electronic device 300 at this predicted time point. In addition, The receiving module 304 can be used as the predicted time point according to the time point set by the user.

The time calculation module 306 is coupled to the receiving module 304. The time calculation module 306 is configured to calculate a time difference between the predicted time point and the current time point. For example, the time calculation module 306 can subtract the predicted time point from the current time point to obtain the time difference.

The determining module 308 is coupled to the time calculating module 306. The determining module 308 compares the time difference and the threshold value obtained by the time calculating module 306 to determine whether the time difference is smaller than the threshold value, wherein the determining module 308 can determine according to the threshold value set by the user. .

The regression analysis module 310 is coupled to the determination module 308 and the recording module 302. In this embodiment, the regression analysis module 310 performs a regression analysis prediction process according to the resource load record recorded by the recording module 302. Specifically, in the regression analysis prediction process, the regression analysis module 310 performs a regression analysis algorithm on the resource load record recorded by the recording module 302 during the time period to obtain a prediction model, thereby predicting the model. The predicted value corresponding to the predicted time point is taken out.

The group analysis module 312 performs a group analysis prediction process based on the resource load record. Specifically, in the group analysis prediction program, the group analysis module 312 selects an average of most of the resource load records as the resource load of the predicted time point in the predicted time segment corresponding to the predicted time point. The predicted value of the record.

The recording module 302, the receiving module 304, the time calculating module 306, the determining module 308, and the regression analyzing module 310 of the embodiment are described herein. And the group analysis module 312 is implemented by a hardware circuit composed of one or several logic gates.

Alternatively, in another embodiment of the present invention, the recording module 302, the receiving module 304, the time calculating module 306, the determining module 308, the regression analyzing module 310, and the group analyzing module 312 may be computer code. Come on. For example, the recording module 302, the receiving module 304, the time calculating module 306, the determining module 308, the regression analyzing module 310, and the group analyzing module 312 are, for example, code fragments written by a programming language. For application, operating system or driver, etc., these code segments are stored in the storage unit and executed by the processing unit.

In addition, in other embodiments, the electronic device 300 further includes a processing unit and a storage unit, and the processing unit is coupled to the recording module 302, the receiving module 304, the time computing module 306, and the determining module. 308. The regression analysis module 310 and the group analysis module 312 drive the modules, and the modules cooperate to complete the functions through the control of the processing unit.

Further, the processing unit is a hardware (for example, a chipset, a processor, or the like) having computing power for controlling the overall operation of the electronic device 300. The processing unit is, for example, a central processing unit (CPU), or other programmable microprocessor (Microprocessor), a digital signal processor (DSP), a programmable controller, and a special application product. Application Specific Integrated Circuits (ASICs), Programmable Logic Devices (PLDs), or other classes Like a device.

In addition, the above storage unit may be an in-line storage unit or an external storage unit. The embedded storage unit can be a random access memory (RAM), a read-only memory (ROM), a flash memory, a magnetic disk storage device (Magnetic disk storage). Device) and so on. The external storage unit can be a Compact Flash (CF) memory card, a Secure Digital (SD) memory card, a Micro SD memory card, a Memory Stick (MS), and the like. In this embodiment, the storage unit may store one or more code codes and materials used to execute the load prediction method (for example, the resource load record, threshold value, and the like recorded by the recording module 302).

In order to enable those skilled in the art to further understand the load prediction method of the present embodiment and the electronic device using the same, the group analysis module 312 will be described in detail below by way of example.

4 is a block diagram of a group analysis module according to a third embodiment of the present invention. Referring to FIG. 4 , the group analysis module 312 includes a partitioning module 402 , a prediction module 404 , a grouping module 406 , a selection module 408 , and a prediction computing module 410 . The functions of each module in the group analysis module 312 are described as follows.

The dividing module 402 is configured to perform the above step S202, that is, divide each time period into a plurality of time segments, so that the resource load record in each time period is divided into a plurality of data segments. The prediction module 404 performs the above step S204, that is, in the time period, the selection is consistent with the predicted time point. Time period. The grouping module 406 is configured to perform the above step S206, that is, the operation of the cluster analysis. The selection module 408 is configured to perform the above step S208, that is, select one of the clusters among the plurality of clusters. The prediction calculation module 410 is configured to perform the above step S210 to calculate an average value of the resource load records.

In addition, the partitioning module 402, the prediction module 404, the grouping module 406, the selection module 408, and the predictive computing module 410 may also be code written by a programming language or a separate chip. A detailed description of the partitioning module 402, the prediction module 404, the clustering module 406, the selection module 408, and the predictive computing module 410 can be referred to FIG. 2 and will not be described in detail herein.

Fourth embodiment

In order to enable those skilled in the art to further understand the load prediction method of the present embodiment and the electronic device using the same, an embodiment will be described below, which will be described with reference to the electronic device 300 of FIG.

In this embodiment, it is assumed that the recording module 302 records one resource time per day and records every resource usage every two minutes, and records 14 electronic devices 300 between May 1 and May 14. Resource load record. That is to say, the recording module 302 records 10080 resource usage amounts in the recorded 14 time periods (indicated by time periods T1 to T14).

Here, it is assumed that the predicted time point received by the receiving module 304 is 9:00 am on May 15th, and the current time point is 8:00 am on May 15th, and the threshold value is 2 hours. Next, the determination module 308 will determine the predicted time point (ie, 9 am on May 15) and the current time point (May 15) The time difference (ie 1 hour) between 8 am) is less than the threshold. Since the judgment result of the judgment module 308 is YES, that is, the time difference between the predicted time point and the current time point is less than the threshold value, the regression analysis module 310 will be based on the electronic data from May 1 to May 14. The regression analysis prediction program is performed by 10080 resource usage of the device 300.

In detail, in the regression analysis prediction program, the regression analysis module 310 performs a regression analysis algorithm based on the 10080 resource usage amount to obtain a prediction model, wherein the prediction model may include the current time point (May 15th) The predicted value corresponding to each time point (ie, every 2 minutes) between the 8:00 am and the predicted time point (ie, 9:00 am on May 15). That is to say, the regression analysis module 310 can bring the prediction time point (ie, 9:00 am on May 15) into the above prediction model, and take the prediction value corresponding to the prediction time point as the prediction model in this prediction model as Forecasted resource usage.

In this embodiment, since the predicted time point is close to the current time point (ie, the time difference is less than the threshold value), the electronic device 300 predicts the resource usage amount through the regression analysis prediction program, thereby effectively predicting the resource usage amount in the future time. The predicted value at the point.

On the other hand, in the above embodiment, if the receiving module 304 receives the predicted time point at 10:00 am on May 15, and the current time point is at 6 am on May 15, and the threshold is 2 hours. . Here, the determination module 308 determines whether the time difference (ie, 3 hours) between the predicted time point (ie, 10 am on May 15) and the current time point (6 am on May 15) is less than the threshold value. Since the judgment result of the judgment module 308 is no, that is, the time difference between the predicted time point and the current time point is not less than the door. Therefore, the group analysis module 312 performs the group analysis prediction process according to the 10080 resource usage of the electronic device 300 between May 1 and May 14.

In detail, in the regression analysis prediction program, the division module 402 divides the time periods T1 T T14 into a plurality of time segments, so that the resource load records in the time periods T1 T T14 are divided into a plurality of data segments.

It should be noted that, in this embodiment, it is assumed that the dividing module 402 divides the time periods T1 T T14 into four time segments (hereinafter represented by time segments TS1 TS TS4 ). That is, each time period (i.e., one day) is divided into four time segments by the partitioning module 402, wherein each of the segments has a time of six hours. More specifically, the time zone TS1 is between 0:00 am and 6:00 am in each time period, and the time zone TS2 is between 6 am and 12 noon in each time period, each one The time zone S3 is between 12 noon and 18 pm in the time period, and the time zone S4 is between 18 pm and 12 pm in each time period.

Next, the prediction module 404 selects one of the predicted time segments corresponding to the predicted time point in the time segments TS1 to TS4. Here, since the predicted time point is 10:00 am on May 15 and corresponds to the time zone S2 of each time period, the prediction module 404 selects the time zone S2 as the predicted time zone (below to predict the time zone) Paragraph Q indicates).

Further, the grouping module 406 performs cluster analysis on the data segments recorded by the recording module 302 in the time period T1~T14 of the prediction time segment Q. Here, it is assumed that the grouping module 406 is divided according to the K-means cluster. The data segment is divided into 5 clusters (under the clusters C1 to C5), and if the cluster C3 has the largest number of data segments, the selection module 408 selects the cluster C3 among the clusters C1 to C5. Next, the prediction calculation module 410 calculates the average value of the data segments included in the time period T1~T14 of the cluster C3 as the predicted value of the predicted time point (ie, 9:00 am on May 15), according to which The predicted value is used as the predicted resource usage.

In this embodiment, since the predicted time point is far from the current time point (ie, the time difference is not less than the threshold value), the electronic device 300 predicts the resource usage amount through the group analysis and prediction program, thereby corresponding to the predicted time point. In the prediction time section, the average of most of the recorded resource usage is selected as the predicted value of the resource usage at the predicted time point. In this way, the electronic device 300 can effectively predict the resource usage amount at a future time point according to the predicted value.

In summary, in the load prediction method and the electronic device of the embodiment of the present invention, the electronic device uses a plurality of resource load records recorded in a plurality of time periods to predict the resource usage amount at the predicted time point. Wherein, when the time difference between the predicted time point and the current time point is less than the threshold, the electronic device predicts the resource usage through the regression analysis prediction program. However, when the time between the predicted time point and the current time point is not less than the threshold, the electronic device predicts the resource usage through the group analysis prediction program. Thereby, the electronic device can predict the resource usage in different application requirements and the operating environment according to the recorded resource load record, to calculate the predicted value of the resource usage amount at the predicted time point. In this way, the electronic device can adjust the resource usage allocation according to the predicted value to avoid The electronic device is unevenly loaded, thereby improving the working efficiency of the electronic device.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the invention, and any one of ordinary skill in the art can make some modifications and refinements without departing from the spirit and scope of the invention. The scope of the invention is defined by the scope of the appended claims.

300‧‧‧Electronic devices

302‧‧‧recording module

304‧‧‧ receiving module

306‧‧‧Time calculation module

308‧‧‧Judgement module

310‧‧‧Regression Analysis Module

312‧‧‧Group Analysis Module

402‧‧‧Division module

404‧‧‧ Prediction Module

406‧‧‧Group Module

408‧‧‧Selection module

410‧‧‧Predictive Computing Module

S102~S112, S202~S210‧‧‧ steps of load prediction method

1 is a flow chart of a load prediction method according to a first embodiment of the present invention.

2 is a flow chart of a group analysis prediction program according to a second embodiment of the present invention.

3 is a block diagram of an electronic device in accordance with a third embodiment of the present invention.

4 is a block diagram of a group analysis module according to a third embodiment of the present invention.

S102~S112‧‧‧Steps of load prediction method

Claims (10)

A load prediction method is applicable to an electronic device, the method comprising: recording a plurality of resource load records of the electronic device in a plurality of time periods; receiving a predicted time point; calculating the predicted time point and a current time point a time difference, wherein the predicted time point is greater than the current time point on a time axis; determining whether the time difference is less than a threshold; if the time difference is less than the threshold, performing a regression analysis prediction process according to the resource load records; And if the time difference is not less than the threshold, a group analysis prediction process is performed according to the resource load records. The load prediction method according to claim 1, wherein the step of performing the regression analysis prediction program according to the resource load records comprises: performing a regression analysis algorithm on the recorded resource load records to obtain a a prediction model; and extracting a predicted value corresponding to the predicted time point in the prediction model. The load prediction method according to claim 1, wherein the step of executing the group analysis prediction program according to the resource load records comprises: dividing each of the time periods into a plurality of time segments, such that each Separating the resource load records into a plurality of data segments; selecting, in the time segments, one of the predicted time segments corresponding to the predicted time points; and the data segments of the predicted time segments in the time periods Performing a cluster analysis to group the pieces of data into a plurality of clusters; from among the clusters, selecting one of the clusters with the largest number of segments; and calculating an average of the pieces of data included in the selected cluster , as the predicted value of the predicted time point. The load prediction method of claim 3, wherein the step of performing the cluster analysis on the pieces of data in the time periods for each of the time segments comprises: calculating each of the time zones A similarity between each of the pieces of data within the segment, and the cluster analysis is performed based on the similarity. The load forecasting method of claim 1, wherein recording the resource load records of the electronic device during the time periods respectively comprises: obtaining, according to a sampling rate, each of the time periods Resource usage, wherein each of the resource load records includes the amount of resource usage obtained during each of the time periods. An electronic device comprising: a recording module for recording a plurality of resource load records of the electronic device in a plurality of time periods; a receiving module receiving a predicted time point; and a time calculating module calculating the predicted time point With a current time a time difference, wherein the predicted time point is greater than the current time point on a time axis; a regression analysis module performs a regression analysis prediction process according to the resource load records; a group analysis module, according to the The resource load record performs a group analysis and prediction process; and a determination module determines whether the time difference is less than a threshold value, and if the time difference is determined to be less than the threshold value, the determining module notifies the regression analysis module to perform the regression analysis The prediction program; if the determination time difference is not less than the threshold, the determining module notifies the group analysis module to execute the group analysis prediction program. The electronic device of claim 6, wherein the regression analysis module performs a regression analysis algorithm on the recorded resource load records to obtain a prediction model, wherein the prediction is taken out in the prediction model. A predicted value corresponding to the time point, wherein the resource load records are recorded by the recording module during the time periods respectively. The electronic device of claim 6, wherein the group analysis module further comprises: a dividing module, dividing each of the time periods into a plurality of time segments, so that each of the resource loads The recording is divided into a plurality of data segments; a prediction module selects one of the predicted time segments corresponding to the predicted time point in the time segments; a segmentation module, wherein the predicted time segment is in the time periods The data segments within the cluster are subjected to a cluster analysis to group the data segments into a plurality of clusters; a selection module from which one of the clusters having the largest number of clusters is selected; and a predictive computing module that calculates an average value of the pieces of data included in the selected cluster to As the predicted value of the predicted time point. The electronic device of claim 8, wherein the grouping module calculates a similarity between each of the pieces of data in each of the time segments, and performs the cluster analysis according to the similarity. The electronic device of claim 6, wherein the recording module acquires a plurality of resource usages according to a sampling rate during each of the time periods, wherein each of the resource load records is included in each The amount of resource usage obtained during these time periods.