A method for scene recommendation in smart homes and smart home devices

Abstract

This invention discloses a smart home scene recommendation method and smart home devices. It generates and stores an original scene usage dataset based on smart home device usage data from different users, environmental data at usage times, and publicly available personal characteristic data from each user. A preset feature filtering algorithm is used to select multiple important features with importance metrics greater than a preset threshold from various basic features. A preprocessed scene usage dataset is then selected from the original scene usage dataset based on these important features. Finally, target scene setting parameters matching the current user are determined from the preprocessed scene usage dataset based on the importance metrics of each important feature, and these target scene setting parameters are pushed to the current user's terminal. This allows for more accurate smart home scene recommendations and improves the user experience.

Description

Technical Field

[0001] This application relates to the field of smart home technology, and more specifically, to a smart home scene recommendation method and a smart home device. Background Technology

[0002] With the rapid development of internet technology and the widespread adoption of smart home appliances in people's daily lives, the concept of smart technology has penetrated into the lives of more and more ordinary consumers. More intelligent use and control of smart home appliances can enhance user experience and increase user loyalty.

[0003] A smart home scene refers to a set of predefined operations triggered when preset execution conditions are met. Currently used scene recommendation methods can only select some scenes from the preset ones to recommend to users, and cannot create new scenes based on user needs, resulting in some recommended scenes not meeting user requirements.

[0004] Therefore, how to more accurately recommend smart home scenarios and improve user experience is a technical problem that needs to be solved. Summary of the Invention

[0005] This invention discloses a scene recommendation method for smart homes, which solves the technical problem of poor accuracy in recommending smart home scenes in the prior art.

[0006] The method includes:

[0007] Original scene usage datasets are generated and stored based on smart home device usage data of different users, environmental data at usage time points, and publicly available personal characteristic data of each user.

[0008] Based on a preset feature selection algorithm, several important features with importance metrics greater than a preset threshold are selected from each basic feature.

[0009] Based on the aforementioned key features, a preprocessed scenario usage dataset is selected from the original scenario usage dataset;

[0010] Based on the importance metric values ​​of each important feature, the target scenario setting parameters matching the current user are determined from the preprocessed scenario usage dataset, and the target scenario setting parameters are pushed to the current user's user terminal.

[0011] Each basic feature is determined based on data from the dataset used in the original scene, excluding scene setting parameters.

[0012] In some embodiments of this application, target scene setting parameters matching the current user are determined from the preprocessed scene usage dataset based on the importance metric values ​​of each important feature. Specifically:

[0013] The target feature data matching the current user are determined in descending order of importance metric values ​​for each important feature;

[0014] Based on the preprocessing scenario, the target scenario setting parameters are determined using one or more scenarios in the dataset that correspond to each target feature data.

[0015] In some embodiments of this application, the target scene setting parameters are determined based on the preprocessing scene using one or more scenes to be recommended in the dataset corresponding to each target feature data, specifically as follows:

[0016] If there is only one scenario to be recommended, the target scenario setting parameters are determined based on the scenario setting parameters of the scenario to be recommended.

[0017] If there are multiple scenarios to be recommended, the target scenario setting parameters are determined based on the last usage time and the number of times each scenario has been used.

[0018] In some embodiments of this application, the target scenario setting parameters are determined based on the last usage time and the number of times each scenario is used, specifically:

[0019] Determine the difference between the last usage time and the current time for each scenario to be recommended, and determine the nearest scenario to be recommended corresponding to the smallest difference among all the differences;

[0020] If there is only one recently recommended scenario, the target scenario setting parameters are determined based on the scenario setting parameters of the recently recommended scenario.

[0021] If there are multiple recently recommended scenarios, the target scenario setting parameters are determined based on the scenario setting parameters of the most frequently used recently recommended scenario.

[0022] In some embodiments of this application, smart home device usage data includes the on / off times and parameter settings of various smart home devices; environmental data includes weather, temperature, humidity, wind speed, and PM2.5; and personal characteristic data includes residential area, gender, age group, and family population composition.

[0023] In some embodiments of this application, the basic features include the user's personal feature data, the type of smart home device, the time point of use of the smart home device, and the environmental data at the time of use.

[0024] In some embodiments of this application, after pushing the target scenario setting parameters to the current user's terminal, the method further includes:

[0025] Obtain the operation result of the current user on the target scene setting parameters from the user terminal;

[0026] If the operation result is to accept the target scene setting parameters, save the scene usage data corresponding to the target scene setting parameters.

[0027] In some embodiments of this application, the preset feature selection algorithm is the random forest algorithm.

[0028] Accordingly, the present invention also proposes a smart home device, comprising:

[0029] The communication module is used to communicate with the user terminal;

[0030] The controller is configured as follows:

[0031] When a control command is received from the user terminal, the smart home devices are controlled according to the control command.

[0032] The control command is triggered by the user terminal after receiving the user's acceptance operation for the target scene setting parameters, and the target scene setting parameters are pushed to the user terminal according to the method described in any one of claims 1-8.

[0033] By applying the above technical solutions, an original scene usage dataset is generated and stored based on the smart home device usage data of different users, environmental data at usage time points, and publicly available personal characteristic data of each user. A preset feature filtering algorithm is used to select several important features from the basic features whose importance metrics are greater than a preset threshold. A preprocessed scene usage dataset is then selected from the original scene usage dataset based on these important features. Target scene setting parameters matching the current user are determined from the preprocessed scene usage dataset based on the importance metrics of each important feature, and these target scene setting parameters are pushed to the current user's terminal. The basic features are determined based on data in the original scene usage dataset other than the scene setting parameters, thus providing more accurate recommendations for smart home scenes and improving the user experience. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating a smart home scene recommendation method proposed in an embodiment of the present invention is shown.

[0036] Figure 2A flowchart illustrating a scene recommendation method for smart homes according to another embodiment of the present invention is shown.

[0037] Figure 3 This diagram illustrates a simple example of determining target feature data in an embodiment of the present invention.

[0038] Figure 4 A schematic diagram of the structure of a smart home device according to an embodiment of the present invention is shown. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0040] This application provides a smart home scene recommendation method, which can be applied to a cloud server, such as... Figure 1 As shown, the method includes the following steps:

[0041] Step S101: Generate and store the original scene usage dataset based on the smart home device usage data of different users, environmental data at usage time points, and publicly disclosed personal characteristic data of each user.

[0042] In this embodiment, different users may have different smart home devices. For example, some users' smart home devices include air conditioners and humidifiers, while others' smart home devices include air conditioners and curtains.

[0043] Based on the log data uploaded by different users' smart home devices, we collect the usage data of different users' smart home devices and the environmental data at the time of use. With the users' knowledge, we collect the publicly available personal characteristic data of each user. The personal characteristic data can be determined by the user's personal profile data. Based on the smart home device usage data, the environmental data at the time of use and the personal characteristic data, we generate and store the original scene usage dataset.

[0044] To make more accurate scene recommendations, in some embodiments of this application, smart home device usage data includes the on / off times and parameter settings of various smart home devices, environmental data includes weather, temperature, humidity, wind speed and PM2.5, and personal characteristic data includes residential area, gender, age group and family population composition.

[0045] The above embodiments are merely one specific implementation scheme proposed in this application. Those skilled in the art can add, subtract, or replace the smart home device usage data, environmental data, and personal characteristic data according to actual needs, which does not affect the protection scope of this application.

[0046] Step S102: Based on the preset feature filtering algorithm, select multiple important features with importance metrics greater than the preset threshold from each basic feature.

[0047] In this embodiment, multiple basic features are determined from the original scene usage dataset based on data other than scene setting parameters. The importance metric of each basic feature can be determined according to a preset feature filtering algorithm. This importance metric represents the degree of importance of the basic feature for scene recommendation. Then, multiple basic features with importance metric values ​​greater than a preset threshold are selected as important features.

[0048] For example, if the basic features have six fields: [region, gender, age group, temperature, humidity, and season], and the important features selected based on the preset feature filtering algorithm are four fields: [region, age group, temperature, and humidity], and the importance metric values ​​w for these four fields are [0.4, 0.3, 0.2, and 0.1], respectively.

[0049] Those skilled in the art can use different preset feature screening algorithms (such as filtering, wrapping, and embedded) to screen according to actual needs, which does not affect the scope of protection of this application.

[0050] To enable more accurate scenario recommendations, in some embodiments of this application, the basic features include the user's personal feature data, the type of smart home device, the usage time of the smart home device, and the environmental data at the usage time.

[0051] The above embodiments are merely one specific implementation scheme proposed in this application. Those skilled in the art can add, subtract, or replace the basic features according to actual needs, which does not affect the protection scope of this application.

[0052] In order to more accurately filter out important features, in some embodiments of this application, the default feature filtering algorithm is the random forest algorithm.

[0053] In machine learning, a random forest is a classifier consisting of multiple decision trees, and its output class is determined by the mode of the classes output by the individual trees. In the random forest algorithm, for each decision tree in the random forest, its out-of-bag (OOB) error is calculated using the corresponding OOB (out-of-bag) data, denoted as errOOB1. Noise is randomly added to the feature X of all samples in the OOB data (this randomly changes the value of the sample at feature X), and its OOB error is calculated again, denoted as errOOB2. Assuming there are N trees in the random forest, the importance of feature X is w = ∑(errOOB2 - errOOB1) / Ntree.

[0054] Step S103: Select a preprocessed scene usage dataset from the original scene usage dataset based on each of the important features.

[0055] In this embodiment, data matching each important feature is selected from the original scene usage dataset to form a preprocessed scene usage dataset.

[0056] Step S104: Determine the target scene setting parameters that match the current user from the preprocessed scene usage dataset based on the importance metric values ​​of each important feature, and push the target scene setting parameters to the current user's user terminal.

[0057] In this embodiment, when the current user uses smart home devices, there are multiple feature data, such as region, gender, age group, temperature, humidity, season, etc. Based on the importance metric value of each important feature, the target scene setting parameters that match the current user are determined from the preprocessed scene usage dataset, and then the target scene setting parameters are pushed to the current user's user terminal.

[0058] User terminals can be mobile phones, tablets, smart wearable devices, or smart home devices with screens specified by the user, such as refrigerators and speakers.

[0059] To achieve more accurate scene recommendation, in some embodiments of this application, target scene setting parameters matching the current user are determined from the preprocessed scene usage dataset based on the importance metrics of each important feature. Specifically:

[0060] The target feature data matching the current user are determined in descending order of importance metric values ​​for each important feature;

[0061] Based on the preprocessing scenario, the target scenario setting parameters are determined using one or more scenarios in the dataset that correspond to each target feature data.

[0062] In this embodiment, the current user is assessed based on the importance metric values ​​of each important feature, ranked from highest to lowest, to determine the target feature data matching the current user. For example, first, the current user's region is determined based on the feature [Region] with w=0.4; then, the age group is determined based on the feature [Age Group] with w=0.3; then, the temperature of the current user's environment is determined based on the feature [Temperature] with w=0.2; and finally, the humidity of the current user's environment is determined based on the feature [Humidity] with w=0.1. Figure 3 The diagram shown is a simplified example of determining the feature data of each target in an embodiment of the present invention.

[0063] Based on the target feature data, one or more scenarios to be recommended can be determined from the preprocessed scenario usage dataset, and the target scenario setting parameters can be determined based on one or more scenarios to be recommended.

[0064] To achieve more accurate scene recommendation, in some embodiments of this application, the target scene setting parameters are determined based on one or more scenes to be recommended in the dataset corresponding to each target feature data, according to the preprocessed scene. Specifically:

[0065] If there is only one scenario to be recommended, the target scenario setting parameters are determined based on the scenario setting parameters of the scenario to be recommended.

[0066] If there are multiple scenarios to be recommended, the target scenario setting parameters are determined based on the last usage time and the number of times each scenario has been used.

[0067] In this embodiment, if there is only one scenario to be recommended, the scenario setting parameters of the scenario to be recommended can be directly used as the target scenario setting parameters. If there are multiple scenarios to be recommended, further selection is required, and the target scenario setting parameters are determined based on the last usage time and the number of times each scenario to be recommended has been used.

[0068] It should be noted that the above embodiments are only one specific implementation scheme proposed in this application. Other methods of determining the target scene setting parameters by using one or more scenes to be recommended in the dataset corresponding to each target feature data according to the preprocessing scene are all within the protection scope of this application.

[0069] To achieve more accurate scene recommendations, in some embodiments of this application, the target scene setting parameters are determined based on the last usage time and the number of times each scene to be recommended has been used. Specifically:

[0070] Determine the difference between the last usage time and the current time for each scenario to be recommended, and determine the nearest scenario to be recommended corresponding to the smallest difference among all the differences;

[0071] If there is only one recently recommended scenario, the target scenario setting parameters are determined based on the scenario setting parameters of the recently recommended scenario.

[0072] If there are multiple recently recommended scenarios, the target scenario setting parameters are determined based on the scenario setting parameters of the most frequently used recently recommended scenario.

[0073] In this embodiment, the difference between the last usage time and the current time of each scenario to be recommended is first determined. The most recently recommended scenario is determined based on the minimum difference. There may be one or more recently recommended scenarios. If there is only one, the scenario setting parameters of the most recently recommended scenario are determined as the target scenario setting parameters. If there are multiple scenarios, the target scenario setting parameters are determined based on the scenario setting parameters of the most recently recommended scenario with the most usage times. This avoids the defect that newly created scenarios are not recommended because they have few usage times.

[0074] To achieve more accurate scene recommendations, in some embodiments of this application, after pushing the target scene setting parameters to the current user's terminal, the method further includes:

[0075] Obtain the operation result of the current user on the target scene setting parameters from the user terminal;

[0076] If the operation result is to accept the target scene setting parameters, save the scene usage data corresponding to the target scene setting parameters.

[0077] In this embodiment, the operation result may include acceptance or rejection. If the user accepts the target scene setting parameters, the scene usage data corresponding to the target scene setting parameters is saved.

[0078] By applying the above technical solutions, an original scene usage dataset is generated and stored based on the smart home device usage data of different users, environmental data at usage time points, and publicly available personal characteristic data of each user. A preset feature filtering algorithm is used to select several important features from the basic features whose importance metrics are greater than a preset threshold. A preprocessed scene usage dataset is then selected from the original scene usage dataset based on these important features. Target scene setting parameters matching the current user are determined from the preprocessed scene usage dataset based on the importance metrics of each important feature, and these target scene setting parameters are pushed to the current user's terminal. The basic features are determined based on data in the original scene usage dataset other than the scene setting parameters, thus providing more accurate recommendations for smart home scenes and improving the user experience.

[0079] To further illustrate the technical concept of this invention, the technical solution of this invention will now be described in conjunction with specific application scenarios.

[0080] This application provides a method for recommending smart home scenarios, such as... Figure 2 As shown, the specific process is as follows:

[0081] 1) Data collection module

[0082] a) Collect smart device usage data from different users (such as the power-on time and parameter settings of air conditioners, the power-on and power-off times of smart lights and smart curtains, etc.) as well as environmental data (weather, temperature, humidity, wind speed, PM2.5, etc.) at the time of use.

[0083] b) Collect publicly available personal information of users (such as profile data including place of residence, gender, age group, and family demographics).

[0084] The data collected by the data collection module forms the original scene usage dataset, and the corresponding data table is shown in Table 1.

[0085] Table 1

[0086]

[0087] 2) Data storage module

[0088] Data acquired by the storage data collection module

[0089] 3) Prediction Model

[0090] Predictive models use feature selection:

[0091] Feature selection algorithms include filtering, wrapper, embedding, and other methods.

[0092] From the data stored in the data storage module, select data other than the parameters set for the usage scenario as the basic features of the model:

[0093] (1) User personal information such as place of residence, gender, age group, and family composition.

[0094] (2) Types of smart devices

[0095] (3) Time points when smart devices are used

[0096] (4) Environmental parameters at the time of use of smart devices

[0097] (5) Other conditions

[0098] Feature selection algorithms are used to filter out important features for the prediction model from basic features, and the importance of each important feature is determined. The following is a brief introduction to the principle of feature selection in the Random Forest model:

[0099] (1) For each decision tree in the random forest, calculate its out-of-bag error using the corresponding out-of-bag data, denoted as errOOB1.

[0100] (2) Randomly add noise interference to the feature X of all samples of the out-of-bag data (which can randomly change the value of the sample at feature X), and calculate its out-of-bag data error again, denoted as errOOB2.

[0101] (3) Assuming there are Ntree trees in the random forest, the importance of feature X is w = ∑(err00B2-errOOB1) / Ntree. If the accuracy of out-of-bag tests drops significantly after adding noise to a feature, it means that the feature has a great impact on the classification results of the sample, that is, its importance is relatively high.

[0102] From the original scenario usage dataset in the data storage module, the selected important features and the corresponding scenario data are retained to form a preprocessed scenario usage dataset.

[0103] For example, the data collection module collects data from six fields: [region, gender, age group, temperature, humidity, and season]. The feature selection module determines that the recommendation model needs four features: [region, age group, temperature, and humidity], and the importance metrics w for these four fields are [0.4, 0.3, 0.2, and 0.1], respectively.

[0104] The data preprocessing tables used in the preprocessing scenario are shown in Table 2.

[0105] Table 2

[0106]

[0107] Based on the features determined in the previous step and the importance w of each feature, the features are judged hierarchically to determine the scene that matches the feature data at the current time point.

[0108] Explanation of layering judgment:

[0109] (1) Determine the current user's location based on the feature [region] with w=0.4;

[0110] (2) Determine the age group of the current user based on the characteristic [age group] with w=0.3;

[0111] (3) Based on the characteristic [temperature] of w=0.2, determine the temperature of the current user's environment;

[0112] (4) Based on the characteristic [humidity] of w=0.1, determine the humidity of the current user's environment.

[0113] For detailed process examples, please refer to Figure 3 .

[0114] If multiple scenarios were identified in the previous step, then the scenarios to be recommended are sorted:

[0115] (1) Sort by the difference between the last time the scene was used and the current time. The smaller the difference, the higher the ranking.

[0116] (2) If there are multiple scenarios with the same difference from the current time point, then sort the scenarios with the same difference from the current time point in descending order of the number of times they are used.

[0117] Finally, the scenario setting parameters of the top-ranked recommended scenario are pushed to the current user's terminal as the target scenario setting parameters.

[0118] The prediction module described above can accurately and in real time create and recommend the necessary scenarios for users based on the current environmental conditions and available smart devices. The sorting operation described above avoids the defect that newly created scenarios are not recommended because they are used infrequently.

[0119] This application also provides a smart home device, such as... Figure 4 As shown, it includes:

[0120] The communication module 100 is used to communicate with the user terminal;

[0121] Controller 200 is configured as follows:

[0122] When a control command is received from a user terminal, the smart home devices are controlled according to the control command.

[0123] The control command is triggered by the user terminal after receiving the user's acceptance operation for the target scene setting parameters, which are pushed to the user terminal according to the method described above.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method of recommending a scene of a smart home, the method comprising: The method comprises: According to the smart home device usage data of different users, the environment data at the usage time point and the personal characteristic data disclosed by each user, an original scene usage data set is generated and stored; Based on a preset feature screening algorithm, important features with importance measure values greater than a preset threshold are screened from each basic feature; According to each important feature, a preprocessed scene usage data set is screened from the original scene usage data set; According to the importance measure values of each important feature, a target scene setting parameter matched with the current user is determined from the preprocessed scene usage data set, and the target scene setting parameter is pushed to the user terminal of the current user; Wherein, each basic feature is determined according to the data in the original scene usage data set except the scene setting parameter; According to the importance measure values of each important feature, a target scene setting parameter matched with the current user is determined from the preprocessed scene usage data set, specifically: According to the importance measure values of each important feature from large to small, each target feature data matched with the current user is determined; According to one or more recommended scenes corresponding to each target feature data in the preprocessed scene usage data set, the target scene setting parameter is determined; According to one or more recommended scenes corresponding to each target feature data in the preprocessed scene usage data set, the target scene setting parameter is determined, specifically: If the recommended scene is one, the target scene setting parameter is determined according to the scene setting parameter of the recommended scene; If the recommended scene is more than one, the target scene setting parameter is determined according to the last use time and the scene usage frequency of each recommended scene.

2. The method of claim 1, wherein, According to the last use time and the scene usage frequency of each recommended scene, the target scene setting parameter is determined, specifically: The difference between the last use time of each recommended scene and the current time point is determined, and the nearest recommended scene corresponding to the minimum difference value is determined; If the nearest recommended scene is one, the target scene setting parameter is determined according to the scene setting parameter of the nearest recommended scene; If the nearest recommended scene is more than one, the target scene setting parameter is determined according to the scene setting parameter of the nearest recommended scene with the most scene usage frequency.

3. The method of claim 1, wherein, The smart home device usage data includes the opening and closing time points and parameter setting values of various types of smart home devices, the environment data includes weather, temperature, humidity, wind speed and PM2.5, and the personal characteristic data includes residence area, gender, age group and family population composition.

4. The method of claim 3, wherein, Each basic feature includes the personal characteristic data of the user, the type of smart home device, the usage time point of smart home device, and the environment data at the usage time point.

5. The method of claim 1, wherein, After pushing the target scene setting parameter to the user terminal of the current user, the method further comprises: Obtaining the operation result of the current user on the target scene setting parameter from the user terminal; If the operation result is to accept the target scene setting parameter, save the scene usage data corresponding to the target scene setting parameter.

6. The method of claim 1, wherein, The preset feature screening algorithm is a random forest algorithm.

7. A smart home device, characterized in that, It comprises: A communication module for communicating with a user terminal; A controller configured to: controlling the smart home device according to the control instruction when the control instruction sent by the user terminal is received; wherein the control instruction is triggered by the user terminal after receiving an acceptance operation of a user on a target scene setting parameter, and the target scene setting parameter is pushed to the user terminal according to the method in any one of claims 1-6.

Images