KR20150081527A - Method and Apparatus for Context-aware Recommendation to Distribute Water in Smart Water Grid - Google Patents

Abstract

Suggested are a method and a device for context-aware recommendation to distribute water resources in a smart water grid. The method for context-aware recommendation to distribute water resources in a smart water grid includes a step of grouping a plurality of end users as a community by considering a water consumption tendency of the end users; a step of recommending a water resource for the end users of the community according to a feedback of a water consumption tendency parameter of the end users and context information; and a step of providing the water resource to the end users without a request from the end users.

Description

TECHNICAL FIELD [0001] The present invention relates to a method and apparatus for context-aware recommendation for a water resource distribution in a smart water grid,

The present invention relates to a recommendation method and apparatus for water resource distribution in smart water grids.

In water utilities all over the world, the water grid is faced with a number of problems due to the severely obsolete infrastructure of some 100 years or so. This causes problems related to loss of water, theft of water, and loss of imports to the facility. Smart Water Grid is a highly efficient next generation water management system to overcome these problems of water resource management system using advanced information and communication technology. Intelligent water management system, which combines communication technology with water resource management system, helps to mitigate regional and temporal imbalance of water resources by precisely controlling water demand, Support the supply of water resources.

In recent years, interest in smart water grids continues to increase, and studies related to monitoring of water distribution systems are increasing. A wireless sensor network based solution for a water grid has been reviewed, where a small sensor can instantly detect a specific event or job condition and pass relevant information to the water management system. According to the reference, the current intelligent wireless sensor system is installed in the water management system to track information on the end user's water consumption activity, preference, location and time. However, most of the solutions currently in use or available on the market do not meet the demand for water supply suitable for the end user in real time. The author of the bibliography outlines water distribution in conjunction with the application of a recommendation system to recognize end user requirements to control the water treatment process. A recommendation system using contextual information such as end user's water consumption preference, water quality, regional form and status of grid network has the potential to deliver more suitable water resources to the end user in real time. The results in a conventional recommendation system can not match the actual situation in which water resources must be dispersed. Context information can impact referrals and support better water allocation for smart water grids. As far as we know, there is no systematic study of the Smart Water Grid that combines the recommendation system and the context awareness of wireless sensor networks.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and apparatus for recommending a context for a water distribution in a smart water grid in consideration of an end user's profile, water type, and network status. We also propose a spectral clustering method for grouping end users into different communities based on the end user's common interests in water resources. We also propose a design method of back propagation (BP) neural network to obtain end user 's preference evaluation list for water resources.

In one aspect, the context recommendation method for distributing water resources in a smart water grid proposed in the present invention includes grouping the end user into a community in consideration of a water consumption propensity of a plurality of end users, Recommending water resources to end users of the community in accordance with the feedback of the user's water consumption propensity parameter, and providing water resources to the end user without the end user's request in accordance with the recommendation.

The clustering of the end user into the community in consideration of the water consumption propensity of the plurality of end users includes generating an adjacent matrix and a gain matrix using the context information and using a spectral clustering algorithm using the adjacent matrix and the gain matrix End users can be clustered.

Recommending water resources to end users of the community in accordance with contextual information and feedback of the end-user's water consumption propensity parameter comprises using a content-based approach to determine a similarity between the end user's water consumption propensity and the water resource characteristics , And estimating the evaluation of the water resources through learning by the back propagation neural network.

Estimating the evaluation of the water resources through learning by the back propagation neural network

Calculates the occurrence probability of each category in the current network state for the unclassified items of the context information, classifies the context information by including the item having the highest occurrence probability in the corresponding category, Can be learned and predicted by neural networks.

According to another aspect of the present invention, there is provided a context recognition recommendation apparatus for distributing water resources in a Smart Water Grid, which includes a community module for grouping the end user into a community in consideration of a water consumption propensity of a plurality of end users, And a recommendation engine module that recommends water resources to end users of the community in accordance with the feedback of the end user's water consumption propensity parameter, a water resource server that provides water resources to the end user, . ≪ / RTI >

According to embodiments of the present invention, the water resource with the highest expected rating is recommended to the end user, and the proposed method can improve the accuracy of the recommendation, along with a better user experience as compared to the existing recommendations.

1 is a schematic diagram of a smart water grid according to one embodiment of the present invention.
FIG. 2 is a flowchart illustrating a method for recommending a context for water resource distribution in a smart water grid according to an exemplary embodiment of the present invention. Referring to FIG.
3 is a diagram illustrating a framework of a context aware recommendation system according to an embodiment of the present invention.
4 is a diagram illustrating a structure of a community network according to an embodiment of the present invention.
FIG. 5 is a view showing a classification category of water resource context information according to an embodiment of the present invention.
6 is a diagram showing a structure of a back propagation neural network according to an embodiment of the present invention.
7 is a block diagram illustrating a configuration of a context recognition recommendation apparatus for distributing water resources in a smart water grid according to an embodiment of the present invention.
8 is a diagram illustrating a result of measuring a learning performance of a designed back propagation neural network according to an embodiment of the present invention.
9 is a diagram illustrating a result of measuring a prediction performance of recommendation according to an embodiment of the present invention.

Context - aware recommendation methods and devices for the proposed distribution of water resources in smart water grids can recommend and distribute water resources for smart water grid networks to end users to recommend perfect water resources to end users in the water grid. Also, by using spectral clustering scheme to cluster end users into each other network community using context information including end user's water consumption preference, water quality, region type and status of grid network, Clustering can be done. In addition, a new recommendation framework for the use of back propagation neural networks can be designed to obtain a list of end user preference ratings for water resources. And, the water resource with the highest expected estimate can be recommended to the end users in the community.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a schematic view of a smart water grid.

As shown in Figure 1, the Smart Water Grid is a highly efficient next generation water management system that uses advanced information and communication technologies to overcome the limitations of this traditional water management system. The Smart Water Grid as a comprehensive water management system is designed for the following applications.

- It uses water resources such as rainwater, recycled water, and sea water.

- Efficiently distributes, manages and transports water to mitigate water imbalances.

- Monitor the stability of water resources network in real time using developed sensor network.

- Carry clean water to the water network.

The existing water resource management systems of the Smart Water Grid have a high water loss due to frequent leakage and the production of water, and the processing cost is high because the water quality is the same regardless of where the water is used and the required quality. And end users in the Smart Water Grid have different interests in water resources. Therefore, a water resource recommendation system is needed to build an intelligent water grid satisfying the interests of various users.

FIG. 2 is a flowchart illustrating a method of recommending a context for a water resource distribution in a smart water grid according to an exemplary embodiment of the present invention. Referring to FIG.

Referring to FIG. 2, the context aware recommendation method for water resource distribution in the smart water grid includes clustering 210 an end user into a community in consideration of the water consumption propensity of a plurality of end users, (220) recommending water resources to end users of the community in accordance with the feedback of the consumption propensity parameter, and providing (230) water resources to the end user without a request of the end user as recommended.

In step 210, the end user may be grouped into a community considering the water consumption propensity of a plurality of end users. At this time, an adjacent matrix and a gain matrix are generated using the context information, and the end user can be clustered by a spectral clustering algorithm using an adjacent matrix and a gain matrix. In other words, depending on the propensity of end users, users with similar interests may need the same kind of water resources. End users can be grouped into different communities considering the user's water consumption propensity. This could be a way to address the needs of different users' water resources.

In step 220, water resources may be recommended to end users of the community according to contextual information and feedback of the end user's water consumption propensity parameter. The step of recommending water resources to end users of the community in accordance with the context information and the feedback of the water consumption propensity parameter of the end user may include calculating a similarity between the water consumption propensity of the end user and the water resource characteristics using a content based approach 221), and predicting the evaluation of the water resources (222) through learning by the back propagation neural network. The step of predicting the evaluation of the water resources through learning by the back propagation neural network can calculate the occurrence probability of each category in the current network state for the items not classified in the context information. Then, the context information can be classified by including the item having the highest probability of occurrence in the category. And, it can be learned by back propagation neural network, and water resource evaluation can be predicted. Will be described in detail with reference to FIG.

Referring to FIG. 3, the recommendation system 310 may recommend water resources to the end user of the community 330 according to the context information 350 and the feedback of the user's water consumption parameters, as shown in FIG. For example, in this module, it is assumed that M kinds of water resources 340 (e.g., fresh water, rainwater, sea water, etc.) are stored in the resource server and L users are distributed within the community 330 . The user may select the water source 340 for the end user based on the end user's profile 320 and the water consumption history. In practice, this is a multidimensional input and multidimensional output model, in which the characteristics r _m of the resource, the context r _c , the user profile r _u, and the state r _t of the smart water grid are input and the scenario o _s , the water source o _m and the score o _sc are output , Which can be defined as in Equation (1).

O (o _s , o _m , o _sc ) = R (r _m , r _c , r _u , r _t )

In step 230, water resources may be provided to the end user without the end user's request, as recommended. For example, M kinds of water resources may be stored in the resource server. These can be located at the edge of the network. Two patterns can work for water distribution in the Smart Water Grid. One of them is that the end user sends a request to the water server, and the water server can deliver the requested water to the user. In another pattern, the recommendation engine recommends water for the end user, so that the water server can unilaterally provide water to the user without the user's request.

3 is a diagram illustrating a framework of a context aware recommendation system according to an embodiment of the present invention.

Referring again to FIG. 3, the context-sensitive water resource recommendation scheme for water resource distribution in the smart water grid will be described in detail.

The proposed invention designs and implements a Smart Water Grid to develop a water recommendation system with high prediction accuracy and low time delay, and recommends an approximate recommendation to the end users belonging to the community using the recommended system 310 can do. The key goal of the proposed invention is to increase the accuracy of recommendations in recommendation schemes. Spectral clustering methods may be employed to cluster the end users into different communities 331, 332, and 333, respectively, based on the end user's common interests with the water resources 340. Here, the other community may include an agricultural area 331, an industrial estate 332, a residential area 333, and the like. Then, the water resource evaluation obtained by the content-based method can be input to the back propagation neural network to predict the end user's addition request.

This recommendation method can satisfy the requirements of in-time and pull-push applications using the context information 350. Describing the context information 350 effectively is very important. For example, an ontology approach may be used to describe the context information 350 model as follows.

- User's behavior context: time, place, activity, etc.

- User's water consumption context: type of water, amount of water, time, purpose, etc.

- Water grid network context: network status, coverage, capacity, power, etc.

The end users can be grouped into different communities 331, 332, and 333, respectively, based on the user's common interests in the water resources. At this time, spectral clustering algorithm can be constructed using spectral graph theory. This has the advantages of clustering in arbitrary type of sample space as compared with the conventional clustering algorithm and convergence to the global optimal solution is good. This new spectral clustering method for discovering communities can be realized by building adjacency matrices and gain matrices. For example, it can be assumed that there are N users in the Smart Water Grid. And we can define the spectral clustering algorithm based on the gain function to cluster N users into K communities.

For example, the adjacency matrix is assumed to have N end users in the network, and Λ is called an adjacency matrix in the network. If there is a relationship between user i and j, this represents a common concern, Λ _ij = 1, and if not, Λ _ij = 0. In order to express a mutual matrix set G, an element of a matrix G can be expressed by Equation (2).

수학식2

Equation 2

Here, Λ _ij and Λ _kj are elements of the adjacency matrix Λ, and if both user i and user j have a connection side with user k, then Λ _ik Λ _ij = 1, which means that between user i and user j The number of common neighbors (common interests).

로 정의할 수 있다. 이득함수는 수학식3과 같이 표현될 수 있다.
For example, the gain matrix can define g _i and g _j as the rank between user i and user j. This rating can indicate the number of connections between users and other users in the network. The number of common neighbors in any pair of users

. The gain function can be expressed as Equation (3).

수학식3

Equation 3

는 공동체 내에서의 회원함수 (membership function)로 정의 한다. 사용자 i와 사용자 j가 동일한 공동체 내에 있으면

=1이 되고, 그렇지 않으면

=0이 된다. Φ는 공통 관심사에 기반한 공통 이웃의 개수와 임의의 관심사에 기반한 공통 이웃의 개수와의 차이로 정의될 수 있다. 공동체의 마크 벡터(mark vector)는 S = (s1, s2 ... sn)로 표현된다. 사용자 i가 제1 공동체에 속해있으면 si = +1이 되고, 그렇지 않다면 si = -1이 된다.

가 되고,
From here,

Is defined as a membership function in the community. If user i and user j are in the same community

= 1, otherwise

= 0. May be defined as the difference between the number of common neighbors based on a common interest and the number of common neighbors based on any interest. The mark vector of the community is expressed as S = (s1, s2 ... sn). If user i belongs to the first community, si = +1, otherwise si = -1.

Lt; / RTI &

수학식4

Equation 4

수학식5

Equation 5

Only contiguous items for the community can be considered. From here,

수학식6

Equation 6

The gain matrix C can be defined as follows.

수학식7

Equation 7

4 is a diagram illustrating a structure of a community network according to an embodiment of the present invention. Based on the adjacency matrix Λ and the gain matrix C defined by the above-described method, a complex network can be grouped into two communities of community A (410) and community B (420), as shown in FIG. The process of clustering into the community is as follows. First, a principal eigenvector can be calculated from the maximum eigenvalues of the gain matrix C. The network based on the sign of the main element in the eigenvector can then be grouped into two communities. Each of these communities can be separated into other communities by a new spectral clustering method. For example, community A 410 may be divided into a plurality of communities 411, 412, 413, 414, 415, and community B 420 may be divided into a plurality of communities 421, 422, 423, 424, 425 < / RTI >

FIG. 5 is a view showing a classification category of water resource context information according to an embodiment of the present invention. 5, the water resource context information 510 may be classified into time 520, location 530, utilization 540, network 550, and the like. This separate category can be further subdivided. For example, time 520 may be subdivided into morning 521, afternoon 522, evening 523, and location 530 may be subdivided into residential complex 531, industrial complex 532, agricultural area 533, . ≪ / RTI > In addition, utilization 540 may be subdivided into rainwater 541, fresh water 542, and reuse water 543, and network 550 may be subdivided into good 551, normal 552, poor 533 .

First, the similarity of the water resources can be calculated to recommend water resources to the end users of the community according to the context information and the feedback of the end user's water consumption propensity parameter. The content-based recommendation system can recommend water resources to end users who have similar interests in water resources in water use history. For example, assuming that there are n parameters describing the end user, the preference weights of user i for the initial characteristics can be expressed as w = (w1, w2, ..., wn). Then, the correlation between the initial characteristics for the water resources can be expressed by the vector c = (c1, c2,,,, cn). The similarity between user propensity and water resource characteristics can be calculated through a content - based approach. The cosine similarity equation is defined as Equation (8).

수학식8

Equation 8

In this way, a similarity eigenvector space S = {s1, s2, ..., sm} can be obtained, which can represent the similarity between user i and M water resources.

Next, the evaluation of water resources can be predicted. The evaluation of these water resources is learned and predicted by back propagation neural networks. In addition, the context information may be classified by a naive Bayes network. And back propagation neural network was designed to predict the evaluation of water resources. The Bayesian network can calculate the probability of occurrence of each category in the current network state for unclassified items. And, the item with the highest probability can be classified as belonging to the category. The Bayesian network can be used to classify context sensitive information as shown in Figure 5. For example, classification of context-aware information into k categories based on the Bayesian network is defined as {λ ₁ , λ ₂ , ..., λ _k }, k = 1, 2, ..., It can be. The probability that the user requests the water source x _j in the context situation? _I can be calculated using Equations 9 to 11 based on the user's history.

P (x _i | x _i ) = P (x _j | λ _i ) P (λ _i ) / P (x _j )

수학식10

Equation 10

수학식11

Equation 11

Here, a {t _v} is the N characteristic of water x _j, r (λ _i) is the number of my request scenario, T is the number of similar requests water and x _j, T (t _v) is a characteristic t _v The number of requests for water resources, T (λ _i ), can indicate the number of times water is requested in the scenario λ _i . The rest may be represented in the same _{way, {p (λ 1 | x} j), p (λ 2 | x j), ..., p (λ k | x j)} in K-dimensional context, represented by the vector Can be obtained.

6 is a diagram showing a structure of a back propagation neural network according to an embodiment of the present invention.

The back propagation neural network may include three layers: an input layer 610, a hidden layer 620, and an output layer 630. There is a great advantage that weights can be adjusted by considering the feedback on the performance of the user in these three layers. The number of input neurons 611, 612, 613 is equal to the number of elements of the context vector [lambda], and probability values p (λ _k | x _j ) in different context situations can be used as inputs to the input layer have. The number of neurons in the hidden layer 620 is not a constant value and may vary according to the mean squared error (MSE) of the prediction result. Since the purpose of the back propagation neural network is to obtain an evaluation of the water resources, only one node is designed for the output layer 630. In addition, the sigmoid function can be adopted as a transaction function. If there are multiple inputs to a single neuron, these inputs can be obtained through the weighted sum of the input values X. The output can be calculated through Equation (12).

수학식12

Equation 12

At this time, if the output value of the output node is different from the expected value, the weight value can be changed. In the back propagation phase of the back propagation network, the error signal of the output layer 630 may propagate forward to modify the weight to a negative gradient direction of the error function network. For example, it is assumed that the amount of weight modification from the hidden node j to the output is w _j (t) at time t, which can be expressed as shown in equation (13).

수학식13

Equation 13

Where w _j (t + 1) = w _j (t) -? W _j (t) is the error function and O '(t) is the expected output. The modified weight from the hidden layer 620 to the output layer 630 can be calculated as shown in Equation (14).

_{w j (t + 1) =} w j (t) - ηw j (t) equation (14)

Here,? Represents a learning correction rate. Similarly, the correlation value w _lj (t) from the input node l to the hidden node j can be obtained, which can be calculated using Equations (15) to (17).

O (t) = [O '(t) - O (t)] {O

수학식16

Equation 16

_{w lj (t + 1) =} w lj (t) - ηw lj (t) Equation (17)

Finally, a prediction estimate for the water source x _j can be calculated using equation (18).

? = w _p Sim + w _c P * (? _i | x _j |) + w _n Equation 18

Where w _p is the weight of the similarity, w _c is the weight of the context prediction probability, and w _n is the weight of the network state (w _p + w _c + w _n = 1). The weights are not fixed, but can be adjusted to account for different requirements and network conditions. Thus, using such a recommendation method, M water resources can be learned and predicted, and an evaluation list of predicted resources indicating the resource recommendation level for the user can be obtained through this method. Also, water resources with the highest predictive rating can be recommended to the user.

7 is a block diagram illustrating a configuration of a context recognition recommendation apparatus for distributing water resources in a smart water grid according to an embodiment of the present invention.

The context awareness recommendation device for water resource distribution in the smart water grid may include a community module 710, a recommendation system 720, and a water resource server 730.

The community module 710 may group the end user into a community in consideration of the water consumption propensity of a plurality of end users. At this time, an adjacent matrix and a gain matrix are generated using the context information, and the end user can be clustered by a spectral clustering algorithm using an adjacent matrix and a gain matrix. In other words, depending on the propensity of end users, users with similar interests may need the same kind of water resources. End users can be grouped into different communities considering the user's water consumption propensity. This could be a way to address the needs of different users' water resources.

Recommendation system 720 may recommend water resources to end users of the community based on contextual information and feedback of the end user's water consumption propensity parameter. The recommendation system 720 may use the content-based approach to calculate the similarity between the end user's water consumption propensity and the water resource characteristics and predict the evaluation of the water resources through learning by the back propagation neural network. At this time, the occurrence probability of each category in the current network state can be calculated for the unclassified items in the context information. Then, the context information can be classified by including the item having the highest probability of occurrence in the category. It can also be trained by back propagation neural networks to predict the evaluation of water resources.

The water resource server 730 can provide water resources to the end user even if there is no request from the end user as recommended. For example, M kinds of water resources may be stored in the resource server. These can be located at the edge of the network. Two patterns can work for water distribution in the Smart Water Grid. One of them is that the end user sends a request to the water server, and the water server can deliver the requested water to the user. In another pattern, the recommendation engine recommends water for the end user, so that the water server can unilaterally provide water to the user without the user's request.

Next, the results of the evaluation using the context recognition recommendation method and apparatus for water resource distribution in the proposed smart water grid will be described.

First, the parameter design will be described. Considering that the user's experimental data is not available on the relevant standard platform and the water community network, the simulation method can be adapted to obtain the experimental data from the Smart Water Grid. Three kinds of context information including resource characteristics, context recognition information (time, place, activity) and water grid network status can be considered. 300 contextual experimental data, 50 users in the community, and 10 types of water resources are given as parameters of the simulation. The user joins the system at the start of the simulation and leaves the system at the end of the simulation. Taking these parameters into consideration, it is possible to design a back propagation neural network including five input layer nodes, seven hidden layer nodes, and one output layer node. In addition, since it is difficult to set a standard for different water resource recommendation systems, it is possible to designate three evaluation criteria and display the quality of the water resource recommendation system. The number of recommended water sources Nr and the number of water sources of interest Ni are used to represent the quality of the recommended model, including the precision Pr and the recall Re, which can be expressed as Equation 19.

Pr = NrnNiNr Equation 19

Criterion 1: The precision value is the ratio of the amount of water recommended by the user to the amount of recommended resource, and is thus given, as shown in equation (20).

수학식20

Equation 20

Criterion 2: Recovery rate is the ratio of the amount of water of interest that is recommended to the amount of collected interest resources,

수학식21

21

Criterion 3: The satisfaction value of the user is calculated as the sum of the volume Q _sp of the water resources, the popularity Q _pp of the water resources, the quality Q _{qp of the} water resources, and the weight S _ci , and is given by Equation (22).

score = Sc1Qsp + Sc2Qpp + Sc3Qqp Equation 22

The precision value indicates the accuracy of the recommendation system. The recall value represents the recommended percentage of water resources of interest and the Q _os value represents the user's satisfaction with the recommendation.

Next, evaluation results will be described. Based on the designed parameters of the water distribution method, first, the values for the precision and the number of times can be calculated. For example, this recommendation method can achieve a user satisfaction of 0.84 when the predicted value is 0.82 and the recall is 0.36. Since each user has a different tendency, the evaluation of the water resources may be an unfair target for the user. The normalized evaluation criterion (NRC) is defined as Equation (23).

수학식23

Equation 23

Here, L represents the number of end users in the water use community, and L = 50.

Table 1 is a table showing the number of NRC versus water types.

<Table 1>

Figs. 8 and 9 are diagrams showing measurement results. Fig.

The performance of the proposed recommendation system is evaluated through simulation for NRC and MSE. Table 1 shows the number of water species for NRC in various ways. The NRC of the proposed system is higher than the NRC of the random recommendation system. Figure 8 shows the scalability and convergence of the proposed system, suggesting that the proposed system has fast convergence and high scalability for the designed backpropagation neural network. Figure 9 shows the prediction performance of the recommendation scheme. It is shown that the proposed method has good performance for water resource estimation in recommending water resources.

Context recognition recommendation method and device for water resource distribution in the proposed smart water grid propose water resource distribution recommendation method based on context awareness information of Smart Water Grid to improve end user satisfaction and water network operation efficiency. A new spectral clustering method has been developed to group users into a water use community in the Smart Water Grid, taking into account the context information and the end user 's profile. We developed a backpropagation neural network for the recommendation of water resources, which showed better performance for water resource estimation. We also evaluated the performance of this approach to determine the accuracy of recommendation in terms of NRC and MSE in different scenarios.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI &gt; or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (5)

Clustering the end user into a community in consideration of the water consumption propensity of a plurality of end users;
Recommending water resources to end users of the community in accordance with context information and feedback of the end user's water consumption propensity parameter; And
Providing the water resource to the end user without the request of the end user according to the recommendation
Lt; / RTI &gt; The method according to claim 1,
The step of clustering the end user into the community in consideration of the water consumption propensity of the plurality of end users comprises:
Generating an adjacency matrix and a gain matrix using the context information, and clustering the end user with a spectral clustering algorithm using the adjacency matrix and the gain matrix
Water distribution method. The method according to claim 1,
Recommending water resources to end users of the community in accordance with context information and feedback of the end user's water consumption propensity parameter comprises:
Calculating a similarity between the end-user's water consumption propensity and the water resource characteristics using a content-based approach; And
Predicting the evaluation of water resources through learning by back propagation neural network
Lt; / RTI &gt; The method of claim 3,
Estimating the evaluation of the water resources through learning by the back propagation neural network
Quot;
The method includes calculating the probability of occurrence of each category in the current network state for the unclassified items in the context information, classifying the context information by including the calculated highest occurrence probability item in the corresponding category, &Lt; / RTI &gt;
Water distribution method. A community module for grouping the end user into a community in consideration of a water consumption propensity of a plurality of end users;
A recommendation system that recommends water resources to end users of the community in accordance with context information and feedback of the end user's water consumption propensity parameter; And
According to the recommendation, a water resource server that provides water resources to the end user without requesting the end user
.

Images