A hair straightener intelligent control method and system

Abstract

The present invention provides a method and system for intelligent control of a hair straightener, and relates to the technical field of data analysis and processing. The method comprises: obtaining historical usage data of a user on a hair straightener, analyzing the historical usage data and clustering to generate multiple control behavior patterns and time context patterns, determining multiple control behavior labels and time context labels and constructing a first label combination matrix, determining multiple sub-feature labels of each type of control behavior feature, updating the first label combination matrix to generate a second label combination matrix; calculating the information entropy of each first label combination in the second label combination matrix, determining multiple second label combinations and third label combinations from the first label combination, constructing a sample set based on the multiple second label combinations and the third label combinations and training to obtain a control prediction model; generating a target control scheme through the control prediction model and operating the hair straightener. The present invention achieves the improvement of the intelligent control level of the hair straightener.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis and processing, and in particular to an intelligent control method and system for a hair straightener. Background Art

[0002] Traditional hair straightener control methods are mostly based on fixed set values, lacking flexibility and personalization, and unable to dynamically adjust control strategies based on actual user needs or changes in the external environment. The experience and effect of hair straighteners are often affected by factors such as user experience, time, and environment.

[0003] Some smart hair straighteners have tried to provide a better user experience through temperature control technology and automated adjustments. For example, by integrating hardware such as temperature sensors and humidity sensors, and adjusting their operating modes with simple algorithms to adapt to different usage needs. However, these methods are mostly limited to feedback control based on real-time sensor data, lacking comprehensive analysis and long-term adaptability of user historical behavior patterns and external environmental data, and insufficient response to the personalized needs of different users, resulting in an inability to make more intelligent and precise control adjustments based on actual usage scenarios. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a hair straightener intelligent control method and system, which analyzes the user's historical usage data to obtain the user's behavior patterns and personalized usage needs, realizes personalized intelligent control of the hair straightener, and improves the user's usage experience.

[0005] To achieve the above object, the present invention provides a first aspect of a hair straightener intelligent control method, comprising:

[0006] Obtaining historical usage data of the user on the hair straightener, including control behavior data and external environment data under multiple usage operations;

[0007] Extracting multiple groups of control behavior feature data from the control behavior data under multiple usage operations, clustering the multiple groups of control behavior feature data to generate multiple control behavior patterns, and determining a control behavior label for each control behavior pattern;

[0008] Extracting multiple groups of time context feature data from the external environment data under multiple usage operations, clustering the multiple groups of time context feature data to generate multiple time context patterns, and determining a time context label for each time context pattern;

[0009] Constructing a first label combination matrix according to multiple control behavior labels and time context labels, performing feature classification on multiple groups of control behavior feature data, determining multiple sub-feature labels for each type of control behavior feature, and updating the first label combination matrix based on the multiple sub-feature labels for each type of control behavior feature to generate a second label combination matrix;

[0010] Calculate the information entropy of each first label combination in the second label combination matrix, determine multiple second label combinations from the first label combinations based on the information entropy, and determine multiple third label combinations after feature reconstruction of the remaining first label combinations, construct a sample set based on the multiple second label combinations and the third label combinations, and obtain a control prediction model through sample set training;

[0011] After receiving the user's demand time data for using the hair straightener, a target time scenario mode is determined based on the demand time data, the demand time data is analyzed based on the target time scenario mode through a control prediction model to generate a target control scheme, and the operation of the hair straightener is controlled based on the target control scheme.

[0012] Preferably, determining a plurality of second label combinations from the first label combinations based on information entropy, and determining a plurality of third label combinations after feature reconstruction of the remaining first label combinations includes:

[0013] Recording multiple first label combinations whose information entropy is lower than a first preset distribution threshold as second label combinations, and for the remaining multiple first label combinations, extracting a statistical feature vector of each first label combination based on the control behavior feature data corresponding to the first label combination;

[0014] Based on the control behavior label and the time context label, the remaining multiple first label combinations are classified into context modes to obtain a set of labels to be reconstructed under each context mode, and a similarity parameter between statistical feature vectors of any two first label combinations in each set of labels to be reconstructed is calculated;

[0015] The multiple first label combinations of each to-be-reconstructed label set are merged based on a preset similarity threshold, and multiple third label combinations are determined from the multiple to-be-reconstructed label sets.

[0016] Preferably, a sample set is constructed based on multiple second label combinations and third label combinations, and a control prediction model is obtained by training the sample set, including:

[0017] Determine a target usage data set under each second tag combination and third tag combination according to the historical usage data, and extract a target time feature and a target control behavior feature from each target usage data set;

[0018] Associating the target time features and target control behavior features of each target usage data set with the time context label to which the target usage data set belongs to generate multiple groups of sample data, and constructing a sample set containing the multiple groups of sample data;

[0019] Multiple groups of associated time context labels and target time features in the sample set are used as the input of the control prediction model, and the target control behavior features corresponding to each group of associated time context labels and target time features are used as the training targets of the control prediction model. The control prediction model is trained through the sample set to obtain a trained control prediction model.

[0020] Preferably, a target time scenario mode is determined based on the demand time data, and a target control scheme is generated by analyzing the demand time data based on the target time scenario mode through a control prediction model, including:

[0021] Determine the first time pattern feature vector according to the demand time data, extract the second time pattern feature vector of each time situation pattern, calculate the distance between the first time pattern feature vector and multiple second time pattern feature vectors respectively, and select the time situation pattern to which the second time pattern feature vector with the closest distance belongs as the target time situation pattern;

[0022] Based on the target time scenario model, a time scenario label is added to the demand time data, the demand time data containing the time scenario label is input into the control prediction model, and the target control scheme corresponding to the demand time data is generated through the control prediction model.

[0023] Preferably, determining a plurality of third label combinations from a plurality of to-be-reconstructed label sets includes:

[0024] The second label combination matrix is ​​updated according to the multiple label sets to be reconstructed to generate a third label combination matrix, the information entropy of each first label combination in the multiple label sets to be reconstructed in the third label combination matrix is ​​calculated, and the multiple first label combinations with information entropy lower than the second preset distribution threshold are recorded as third label combinations.

[0025] Preferably, determining a target usage data set under each second tag combination and third tag combination according to the historical usage data, and extracting a target time feature and a target control behavior feature from each target usage data set includes:

[0026] According to the control behavior label and the time context label corresponding to each second label combination and the third label combination, and the multiple sub-feature labels contained in the control behavior label, multiple groups of control behavior feature data and time context feature data that meet the multiple labels are extracted to respectively constitute the target usage data set of each second label combination and the third label combination;

[0027] Based on multiple groups of time context feature data and control behavior feature data in the target usage data set, the feature mean of each time context feature and the feature mean of each sub-feature label are calculated to generate a target time feature of the target usage data set containing feature means corresponding to multiple time context features, and a target control behavior feature containing feature means corresponding to multiple sub-feature labels.

[0028] A second aspect of the present invention provides a hair straightener intelligent control system, which is used to implement the above-mentioned hair straightener intelligent control method, comprising:

[0029] A data collection module, used to obtain the user's historical usage data on the hair straightener, including control behavior data and external environment data under multiple usage operations;

[0030] A control behavior analysis module is used to extract multiple groups of control behavior feature data from the control behavior data under multiple usage operations, cluster the multiple groups of control behavior feature data to generate multiple control behavior patterns, and determine the control behavior label of each control behavior pattern;

[0031] A time context analysis module is used to extract multiple groups of time context feature data from the external environment data under multiple usage operations, cluster the multiple groups of time context feature data to generate multiple time context patterns, and determine the time context label of each time context pattern;

[0032] A label combination analysis module is used to construct a first label combination matrix according to multiple control behavior labels and time context labels, perform feature classification on multiple groups of control behavior feature data, determine multiple sub-feature labels of each type of control behavior feature, and update the first label combination matrix based on the multiple sub-feature labels of each type of control behavior feature to generate a second label combination matrix;

[0033] A label combination reconstruction module is used to calculate the information entropy of each first label combination in the second label combination matrix, determine a plurality of second label combinations from the first label combinations based on the information entropy, and determine a plurality of third label combinations after feature reconstruction of the remaining first label combinations;

[0034] A prediction model training module, used for constructing a sample set based on multiple second label combinations and third label combinations, and obtaining a control prediction model through sample set training;

[0035] The control scheme generation module is used to determine the target time scenario mode based on the demand time data after receiving the user's demand time data for using the hair straightener, analyze the demand time data based on the target time scenario mode through the control prediction model to generate a target control scheme, and control the operation of the hair straightener based on the target control scheme.

[0036] Preferably, the control scheme generation module determines the target time scenario mode based on the demand time data, and generates the target control scheme by analyzing the demand time data based on the target time scenario mode through the control prediction model, including:

[0037] Determine the first time pattern feature vector according to the demand time data, extract the second time pattern feature vector of each time situation pattern, calculate the distance between the first time pattern feature vector and multiple second time pattern feature vectors respectively, and select the time situation pattern to which the second time pattern feature vector with the closest distance belongs as the target time situation pattern;

[0038] Based on the target time scenario model, a time scenario label is added to the demand time data, the demand time data containing the time scenario label is input into the control prediction model, and the target control scheme corresponding to the demand time data is generated through the control prediction model.

[0039] The present invention has the following beneficial effects:

[0040] The present invention extracts and clusters multi-level features of the user's control behavior data and external environment data, determines multiple control behavior labels and time context labels, and constructs a label combination matrix. Then, the information entropy of each label combination is calculated to measure the stability and representativeness of the combination. Combinations with low information entropy are screened as typical patterns, and combinations with high information entropy are analyzed and reconstructed based on statistical characteristics, which effectively reduces the redundancy of labels, generates representative label combinations, constructs a sample set and trains a control prediction model, determines the time pattern feature vector according to the user's demand time data, performs feature matching with historical time context patterns, finds the time context pattern closest to the user's demand, accurately identifies the user's demand in the current usage scenario, and generates automated and personalized control solutions for the user in different scenarios through the control prediction model, so as to help the user obtain the best hair straightening effect in different times and usage scenarios, thereby greatly improving the intelligent control level and user experience of the hair straightener. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic flow chart of a method for intelligently controlling a hair straightener provided by an embodiment of the present invention.

[0042] Figure 2 A schematic structural diagram of an intelligent control system for a hair straightener provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make those skilled in the art better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0044] Figure 1 A schematic diagram showing a flow chart of a hair straightener intelligent control method provided by an embodiment of the present invention is shown in FIG. Figure 1 , the method comprises the steps of:

[0045] Step S1, obtaining historical usage data of the user on the hair straightener, including control behavior data and external environment data under multiple usage operations.

[0046] In this embodiment, a large amount of historical usage data of hair straighteners by users is collected, and these data include but are not limited to control behavior data and external environment data of users under different usage operations. Specifically, the control behavior data include control parameters such as output temperature, heating time, and device pressure set by the user when using the hair straightener. These data can reflect the specific operation behavior of the user and characterize the control behavior of the user regarding the hair straightener. The external environment data mainly involves the time information involved in the process of the user using the hair straightener, such as specific different time periods of the day, weekdays, seasons, special dates, etc. These data reflect the time requirements of the user for using the hair straightener, characterize the time context involved in the user's use of the hair straightener, and can provide background information for analyzing the user's control behavior.

[0047] Step S2: extract multiple groups of control behavior feature data from the control behavior data under multiple usage operations, cluster the multiple groups of control behavior feature data to generate multiple control behavior patterns, determine the control behavior label of each control behavior pattern, extract multiple groups of time context feature data from the external environment data under multiple usage operations, cluster the multiple groups of time context feature data to generate multiple time context patterns, and determine the time context label of each time context pattern.

[0048] In this embodiment, multiple groups of control behavior feature data and multiple groups of time context feature data are extracted from the collected historical data, and these feature data are clustered and analyzed to generate multiple control behavior patterns and time context patterns. Specifically, the control behavior data under each use operation is analyzed, and multiple control behavior features in each group of control behavior data are extracted, including output temperature features, heating time features, equipment pressure features and other control behavior feature data. Cluster analysis is performed on multiple groups of control behavior feature data through a clustering algorithm to generate multiple clusters. In this embodiment, the DBSCAN algorithm is used as an example to cluster feature data. Each cluster obtained by clustering can represent a control behavior pattern. For example, by analyzing the data in each cluster, a rapid heating mode, a low temperature hair care mode, etc. may be obtained. Each control behavior pattern corresponds to a specific set of control behaviors, and a corresponding control behavior label can be added for each control behavior pattern.

[0049] Similarly, for external environment data, multiple time context features are extracted from different data to obtain time context feature data for each set of external environment data, so as to reflect the time changes when the user uses the hair straightener. Multiple clusters are obtained by clustering multiple sets of time context feature data. Each cluster can represent a time context mode, and a corresponding time context label can be added to each time context mode. For example, there may be weekday morning contexts, holiday styling contexts, etc., to reflect the user's demand for the use of hair straighteners at different times.

[0050] Step S3: construct a first label combination matrix according to multiple control behavior labels and time context labels, perform feature classification on multiple groups of control behavior feature data, determine multiple sub-feature labels for each type of control behavior feature, and update the first label combination matrix based on the multiple sub-feature labels for each type of control behavior feature to generate a second label combination matrix.

[0051] In this embodiment, based on the control behavior labels and time context labels obtained in the above steps, combined with the analysis of historical usage data, a first label combination matrix can be constructed, and the first label combination matrix involves a combination of multiple control behavior labels and time context labels, as well as the frequency of occurrence of each combination of control behavior labels and time context labels. Exemplarily, after the historical usage data is classified from the control behavior level and the time context level, for each use operation, the control behavior data and external environment data, or for this use operation, a specific control behavior label and time context label can be obtained. After analyzing the historical usage data, the frequency of occurrence of each combination of control behavior labels and time context labels can be statistically obtained, thereby completing the construction of the first label combination matrix.

[0052] On this basis, multiple groups of control behavior feature data are further characterized and classified, wherein the feature classification process can be divided based on the value range of each control behavior feature, or based on the inherent mode of the hair straightener itself. For example, for heating time, it can be divided into three categories: short, medium, and long, and the equipment pressure can be divided into three categories: low, medium, and high. Finally, each type of control behavior feature can obtain multiple sub-feature labels. After completing the feature classification of the control behavior feature data, more second label combinations can be obtained to form a second label combination matrix, each of which is recorded as a first label combination. Exemplarily, one of the first label combinations includes a control behavior label and a time context label, as well as multiple sub-feature labels. Similarly, after analyzing the historical usage data, the frequency of occurrence of each first label combination in the second label combination matrix in the historical usage data can be obtained, and finally the first label combination matrix can be updated to obtain the second label combination matrix.

[0053] Step S4, calculate the information entropy of each first label combination in the second label combination matrix, determine multiple second label combinations from the first label combinations based on the information entropy, and determine multiple third label combinations after feature reconstruction of the remaining first label combinations, construct a sample set based on the multiple second label combinations and the third label combinations, and obtain a control prediction model through sample set training.

[0054] In this embodiment, based on the frequency of occurrence corresponding to the first label combination, the information entropy is calculated for each first label combination in the second label combination matrix, and the information entropy is used to measure the randomness or uncertainty of the label combination. That is, it reflects the control behavior characteristics and time context characteristics under each combination, as well as the degree of discreteness between multiple sub-features. If the information entropy is low, it means that the characteristics under the combination are more concentrated and regular, and may have a higher representativeness. According to the information entropy analysis results of each first label combination, multiple second label combinations are determined from multiple first label combinations according to a pre-set threshold. For the remaining first label combinations, the historical usage data involved is analyzed, the labels are aggregated and compressed, and the remaining multiple first label combinations are feature reconstructed to generate multiple third label combinations. The third label combination forms a more simplified category by merging similar labels, thereby reducing the number of labels and improving the representativeness of the label combination.

[0055] After obtaining a plurality of more representative second label combinations and third label combinations, relevant feature data is extracted from the historical usage data corresponding to each second label combination and third label combination, and a sample set for training a control prediction model is constructed, which is used to train a control prediction model that can be used to predict the control scheme of a hair straightener.

[0056] Step S5: after receiving the time data required for the user to use the hair straightener, determine the target time scenario mode based on the time data required, analyze the time data required based on the target time scenario mode through the control prediction model to generate a target control scheme, and control the operation of the hair straightener based on the target control scheme.

[0057] In this embodiment, when the user uses the hair straightener or needs to use it, real-time analysis can be performed based on the received demand time data, and a time scenario pattern that matches the demand time data can be determined and recorded as a target time scenario pattern. The demand time data can be analyzed by combining the control prediction model with the target time scenario pattern, and finally a target control scheme corresponding to the demand time data is generated, which includes corresponding equipment control parameters such as temperature, usage time, and pressure. The hair straightener can then perform corresponding operation control to achieve personalized intelligent adjustment of the hair straightener.

[0058] Optionally, in one implementation process of the above step S4, multiple second label combinations are determined from the first label combinations based on information entropy, and multiple third label combinations are determined after feature reconstruction of the remaining first label combinations, specifically including:

[0059] Multiple first label combinations whose information entropy is lower than a first preset distribution threshold are recorded as second label combinations, and for the remaining multiple first label combinations, a statistical feature vector of each first label combination is extracted based on the control behavior feature data corresponding to the first label combination.

[0060] Among them, the smaller the information entropy, the smaller the variability of the corresponding first label combination in the data set, which represents a more common or stable pattern. Through the first preset distribution threshold screening, some representative combinations are screened out from multiple first label combinations and recorded as second label combinations. For the remaining first label combinations, the control behavior characteristic data corresponding to each combination are further analyzed to determine multiple statistical features in multiple groups of control behavior characteristic data, such as mean, standard deviation, extreme value and other characteristic parameters, to characterize the central tendency and volatility of the data under the label combination, and construct a statistical feature vector for each first label combination based on the extracted multiple statistical features.

[0061] Based on the control behavior label and the time context label, the remaining multiple first label combinations are classified into context modes to obtain a set of labels to be reconstructed under each context mode, and the similarity parameters between the statistical feature vectors of any two first label combinations in each set of labels to be reconstructed are calculated.

[0062] Among them, the situational pattern classification is based on the control behavior label and time context label of the first label combination, and multiple first label combinations with the same control behavior label and time context label are classified into the same type. In this way, a set of labels to be reconstructed under each situational pattern can be obtained, wherein the situational patterns represent a combination of a control behavior label and a time context label, and the set of labels to be reconstructed contains control behavior feature data and time context feature data corresponding to multiple first label combinations with the same situational patterns. For multiple sets of labels to be reconstructed, the similarity between each first label combination is calculated to obtain a similarity parameter between any two first label combinations. Specifically, a similarity analysis is performed on the statistical feature vectors between the combinations, for example, the similarity parameter is calculated based on the cosine similarity, and the absolute value of the cosine similarity is taken as the similarity parameter. The larger the similarity parameter, the higher the degree of association between the two combinations, which serves as the basis for subsequent feature reconstruction operations.

[0063] Then, based on a preset similarity threshold, multiple first label combinations of each label set to be reconstructed are merged. Specifically, first label combinations whose similarity parameters are greater than the preset similarity threshold are merged to obtain a new first label combination, thereby realizing a merged update of multiple label sets to be reconstructed, and finally determining multiple third label combinations from the multiple updated label sets to be reconstructed.

[0064] In this embodiment, a third label combination matrix is ​​constructed based on multiple label sets to be reconstructed, including merging the relevant label combinations according to the above-mentioned merging operation on the basis of the second label combination matrix and modifying the corresponding occurrence frequency. After the second label combination matrix is ​​updated to obtain the third label combination matrix, the information entropy of each first label combination in the multiple label sets to be reconstructed is calculated in the third label combination matrix, and then the first label combinations with multiple information entropies lower than the second preset distribution threshold in the multiple label sets to be reconstructed are recorded as third label combinations. Among them, the value of the second preset distribution threshold is less than the third preset distribution threshold, and the specific data of the threshold can be reasonably set based on the actual situation of the data, and it is not specifically limited in this embodiment. In this way, based on the control behavior label and the time context label as the merging basis, some label combinations that are relatively similar in the detailed control of the hair straightener are merged, and finally a plurality of representative second label combinations and third label combinations are obtained.

[0065] Optionally, in one implementation process of the above step S4, a sample set is constructed based on multiple second label combinations and third label combinations, and a control prediction model is obtained by training the sample set, specifically including:

[0066] A target usage data set under each second tag combination and third tag combination is determined according to the historical usage data, and a target time feature and a target control behavior feature are extracted from each target usage data set.

[0067] Among them, the target usage data set includes the historical usage data involved in the corresponding label combination, which can be specifically the corresponding control behavior feature data and time context feature data. Through the above steps, the historical usage data is refined and multiple representative label combinations are extracted. The control behavior feature data and time context feature data under each label combination have relatively concentrated feature information. After merging the control behavior feature data and time context feature data in the target usage data set, representative feature values ​​of each type of time feature and control behavior feature can be obtained. Specifically, the feature mean of each time context feature and the feature mean of each sub-feature label are calculated to generate the target time feature of the target usage data set containing the feature mean corresponding to multiple time context features, and the target control behavior feature containing the feature mean corresponding to multiple sub-feature labels.

[0068] Exemplarily, for features such as equipment pressure, the average of multiple equipment pressure features in the target usage data set is taken as the representative feature value of this type of feature. The target time feature and target control behavior feature finally generated represent stable feature information, such as "Sunday afternoon in March", "heating for 15 minutes at low pressure and low temperature", and other specific and stable time information and control behavior information.

[0069] After extracting the target time features and target control behavior features of each target usage data set, the target time features and target control behavior features of each target usage data set are associated with the time context labels to which the target usage data set belongs to generate multiple groups of sample data. Finally, a sample set containing multiple groups of sample data is constructed, which can be used to train the control prediction model.

[0070] In the process of training the control prediction model through the sample set, multiple groups of associated time context labels and target time features in the sample set are used as the input of the control prediction model, and the target control behavior features corresponding to each group of associated time context labels and target time features are used as the training targets of the control prediction model. In this way, the control prediction model is trained to learn the mapping relationship between the input time context labels and target time features and the corresponding target control behavior features, so that the model finally trained can predict the corresponding control scheme according to the input time-related data.

[0071] In this embodiment, a random forest model is specifically selected to construct the control prediction model. The random forest model captures complex feature associations through a combination of decision trees, and can be used to capture the correlation information between complex time features and control behavior features, thereby achieving accurate prediction of control schemes.

[0072] Optionally, in one of the implementation processes of the above step S5, a target time scenario mode is determined based on the demand time data, and a target control scheme is generated by analyzing the demand time data based on the target time scenario mode through a control prediction model, specifically including:

[0073] Determine the first time pattern feature vector according to the demand time data, extract the second time pattern feature vector of each time context pattern, calculate the distances between the first time pattern feature vector and multiple second time pattern feature vectors respectively, and select the time context pattern to which the second time pattern feature vector with the closest distance belongs as the target time context pattern.

[0074] In this embodiment, multiple time context features are extracted from the collected demand time data, and a first time pattern feature vector is constructed based on the multiple time context features. For the multiple time context patterns obtained by clustering multiple groups of time context feature data in the above steps, multiple representative time context features of each time context pattern are extracted, and a second time pattern feature vector of each time context pattern is constructed, and then the distance between the first time pattern feature vector and the multiple second time pattern feature vectors is calculated respectively to match the demand time data with the multiple time context patterns, for example, by measuring the distance between the first time pattern feature vector and the multiple second time pattern feature vectors according to the Euclidean distance, and finally the time context pattern to which the second time pattern feature vector closest to the first time pattern feature vector belongs is recorded as the target time context pattern, and the time context pattern that best matches the demand time data is obtained.

[0075] Then, based on the target time scenario pattern, a time scenario label is added to the demand time data, and finally the demand time data containing the time scenario label is input into the control prediction model. After the time scenario label and the specific time-related characteristics under the time scenario label are known, the model can infer the control parameters that the user may need in a specific time scenario based on the association between the time scenario pattern and the control behavior in the historical data, thereby generating a target control scheme corresponding to the demand time data. Finally, based on the target control scheme, the hair straightener can be intelligently controlled under the demand time data to meet the user's personalized needs and enhance the user's experience of using the hair straightener.

[0076] The present invention provides an intelligent control method for a hair straightener. After determining a plurality of control behavior labels and time context labels and constructing a label combination matrix through multi-level feature extraction and cluster analysis of user control behavior data and external environment data, the stability and representativeness of the combination are measured by calculating the information entropy of each label combination. Combinations with low information entropy are screened as typical patterns, and combinations with high information entropy are subjected to similarity analysis and reconstruction based on statistical characteristics, thereby effectively reducing label redundancy, generating representative label combinations, constructing a sample set and training a control prediction model, determining a time pattern feature vector according to user demand time data, and performing feature matching with historical time context patterns to find a time context pattern closest to the user demand, accurately identifying user demand in the current use scenario, and generating automated and personalized control solutions for users in different scenarios through the control prediction model, thereby helping users to obtain the best hair straightening effect in different times and use scenarios, and greatly improving the intelligent control level and user experience of the hair straightener.

[0077] Figure 2 FIG. 1 is a schematic diagram showing a structure of an intelligent control system for a hair straightener provided by an embodiment of the present invention. Figure 2 The system comprises:

[0078] A data collection module, used to obtain the user's historical usage data on the hair straightener, including control behavior data and external environment data under multiple usage operations;

[0079] A control behavior analysis module is used to extract multiple groups of control behavior feature data from the control behavior data under multiple usage operations, cluster the multiple groups of control behavior feature data to generate multiple control behavior patterns, and determine the control behavior label of each control behavior pattern;

[0080] A time context analysis module is used to extract multiple groups of time context feature data from the external environment data under multiple usage operations, cluster the multiple groups of time context feature data to generate multiple time context patterns, and determine the time context label of each time context pattern;

[0081] A label combination analysis module is used to construct a first label combination matrix according to multiple control behavior labels and time context labels, perform feature classification on multiple groups of control behavior feature data, determine multiple sub-feature labels of each type of control behavior feature, and update the first label combination matrix based on the multiple sub-feature labels of each type of control behavior feature to generate a second label combination matrix;

[0082] A label combination reconstruction module is used to calculate the information entropy of each first label combination in the second label combination matrix, determine a plurality of second label combinations from the first label combinations based on the information entropy, and determine a plurality of third label combinations after feature reconstruction of the remaining first label combinations;

[0083] A prediction model training module, used for constructing a sample set based on multiple second label combinations and third label combinations, and obtaining a control prediction model through sample set training;

[0084] A control scheme generating module, for determining a target time scenario mode based on the demand time data after receiving the demand time data of the user using the hair straightener, analyzing the demand time data based on the target time scenario mode through a control prediction model to generate a target control scheme, and controlling the operation of the hair straightener based on the target control scheme;

[0085] Specifically, a first time pattern feature vector is determined according to the demand time data, a second time pattern feature vector of each time context mode is extracted, the distances between the first time pattern feature vector and multiple second time pattern feature vectors are calculated respectively, and the time context mode to which the second time pattern feature vector with the closest distance belongs is selected and recorded as the target time context mode;

[0086] Based on the target time scenario model, a time scenario label is added to the demand time data, the demand time data containing the time scenario label is input into the control prediction model, and the target control scheme corresponding to the demand time data is generated through the control prediction model.

[0087] The above-mentioned intelligent control system for a hair straightener is specifically used to implement the above-mentioned intelligent control method for a hair straightener during operation.

[0088] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A hair straightener intelligent control method, characterized in that: include: Obtaining historical usage data of the user on the hair straightener, including control behavior data and external environment data under multiple usage operations; Extracting multiple groups of control behavior feature data from the control behavior data under multiple usage operations, clustering the multiple groups of control behavior feature data to generate multiple control behavior patterns, and determining a control behavior label for each control behavior pattern; Extracting multiple groups of time context feature data from the external environment data under multiple usage operations, clustering the multiple groups of time context feature data to generate multiple time context patterns, and determining a time context label for each time context pattern; Constructing a first label combination matrix according to multiple control behavior labels and time context labels, performing feature classification on multiple groups of control behavior feature data, determining multiple sub-feature labels for each type of control behavior feature, and updating the first label combination matrix based on the multiple sub-feature labels for each type of control behavior feature to generate a second label combination matrix; Calculate the information entropy of each first label combination in the second label combination matrix, determine multiple second label combinations from the first label combinations based on the information entropy, and determine multiple third label combinations after feature reconstruction of the remaining first label combinations, construct a sample set based on the multiple second label combinations and the third label combinations, and obtain a control prediction model through sample set training; After receiving the user's demand time data for using the hair straightener, a target time scenario mode is determined based on the demand time data, the demand time data is analyzed based on the target time scenario mode through a control prediction model to generate a target control scheme, and the operation of the hair straightener is controlled based on the target control scheme.

2. The intelligent control method for a hair straightener according to claim 1, characterized in that: Determining a plurality of second label combinations from the first label combinations based on information entropy, and determining a plurality of third label combinations after feature reconstruction of the remaining first label combinations, including: Recording multiple first label combinations whose information entropy is lower than a first preset distribution threshold as second label combinations, and for the remaining multiple first label combinations, extracting a statistical feature vector of each first label combination based on the control behavior feature data corresponding to the first label combination; Based on the control behavior label and the time context label, the remaining multiple first label combinations are classified into context modes to obtain a set of labels to be reconstructed under each context mode, and a similarity parameter between statistical feature vectors of any two first label combinations in each set of labels to be reconstructed is calculated; The multiple first label combinations of each to-be-reconstructed label set are merged based on a preset similarity threshold, and multiple third label combinations are determined from the multiple to-be-reconstructed label sets.

3. The intelligent control method for a hair straightener according to claim 2, characterized in that: A sample set is constructed based on multiple second label combinations and third label combinations, and a control prediction model is obtained through sample set training, including: Determine a target usage data set under each second tag combination and third tag combination according to the historical usage data, and extract a target time feature and a target control behavior feature from each target usage data set; Associating the target time features and target control behavior features of each target usage data set with the time context label to which the target usage data set belongs to generate multiple groups of sample data, and constructing a sample set containing the multiple groups of sample data; Multiple groups of associated time context labels and target time features in the sample set are used as the input of the control prediction model, and the target control behavior features corresponding to each group of associated time context labels and target time features are used as the training targets of the control prediction model. The control prediction model is trained through the sample set to obtain a trained control prediction model.

4. The method for intelligently controlling a hair straightener according to claim 3, characterized in that: Determine the target time scenario mode based on the demand time data, and generate a target control plan by analyzing the demand time data based on the target time scenario mode through the control prediction model, including: Determine the first time pattern feature vector according to the demand time data, extract the second time pattern feature vector of each time situation pattern, calculate the distance between the first time pattern feature vector and multiple second time pattern feature vectors respectively, and select the time situation pattern to which the second time pattern feature vector with the closest distance belongs as the target time situation pattern; Based on the target time scenario model, a time scenario label is added to the demand time data, the demand time data containing the time scenario label is input into the control prediction model, and the target control scheme corresponding to the demand time data is generated through the control prediction model.

5. The intelligent control method for a hair straightener according to claim 2, characterized in that: Determining a plurality of third label combinations from a plurality of to-be-reconstructed label sets includes: The second label combination matrix is ​​updated according to the multiple label sets to be reconstructed to generate a third label combination matrix, the information entropy of each first label combination in the multiple label sets to be reconstructed in the third label combination matrix is ​​calculated, and the multiple first label combinations with information entropy lower than the second preset distribution threshold are recorded as third label combinations.

6. A hair straightener intelligent control method according to claim 5, characterized in that: Determine a target usage data set under each second label combination and third label combination according to the historical usage data, and extract a target time feature and a target control behavior feature from each target usage data set, including: According to the control behavior label and the time context label corresponding to each second label combination and the third label combination, and the multiple sub-feature labels contained in the control behavior label, multiple groups of control behavior feature data and time context feature data that meet the multiple labels are extracted to respectively constitute the target usage data set of each second label combination and the third label combination; Based on multiple groups of time context feature data and control behavior feature data in the target usage data set, the feature mean of each time context feature and the feature mean of each sub-feature label are calculated to generate a target time feature of the target usage data set containing feature means corresponding to multiple time context features, and a target control behavior feature containing feature means corresponding to multiple sub-feature labels.

7. An intelligent control system for a hair straightener, characterized in that: The system is used to implement the hair straightener intelligent control method according to any one of claims 1 to 6, comprising: A data collection module, used to obtain the user's historical usage data on the hair straightener, including control behavior data and external environment data under multiple usage operations; A control behavior analysis module is used to extract multiple groups of control behavior feature data from the control behavior data under multiple usage operations, cluster the multiple groups of control behavior feature data to generate multiple control behavior patterns, and determine the control behavior label of each control behavior pattern; A time context analysis module is used to extract multiple groups of time context feature data from the external environment data under multiple usage operations, cluster the multiple groups of time context feature data to generate multiple time context patterns, and determine the time context label of each time context pattern; A label combination analysis module is used to construct a first label combination matrix according to multiple control behavior labels and time context labels, perform feature classification on multiple groups of control behavior feature data, determine multiple sub-feature labels of each type of control behavior feature, and update the first label combination matrix based on the multiple sub-feature labels of each type of control behavior feature to generate a second label combination matrix; A label combination reconstruction module is used to calculate the information entropy of each first label combination in the second label combination matrix, determine a plurality of second label combinations from the first label combinations based on the information entropy, and determine a plurality of third label combinations after feature reconstruction of the remaining first label combinations; A prediction model training module, used for constructing a sample set based on multiple second label combinations and third label combinations, and obtaining a control prediction model through sample set training; The control scheme generation module is used to determine the target time scenario mode based on the demand time data after receiving the user's demand time data for using the hair straightener, analyze the demand time data based on the target time scenario mode through the control prediction model to generate a target control scheme, and control the operation of the hair straightener based on the target control scheme.

8. The intelligent control system for a hair straightener according to claim 7, characterized in that: For the control scheme generation module, the target time scenario mode is determined based on the demand time data, and the demand time data is analyzed based on the target time scenario mode through the control prediction model to generate a target control scheme, including: Determine the first time pattern feature vector according to the demand time data, extract the second time pattern feature vector of each time situation pattern, calculate the distance between the first time pattern feature vector and multiple second time pattern feature vectors respectively, and select the time situation pattern to which the second time pattern feature vector with the closest distance belongs as the target time situation pattern; Based on the target time scenario model, a time scenario label is added to the demand time data, the demand time data containing the time scenario label is input into the control prediction model, and the target control scheme corresponding to the demand time data is generated through the control prediction model.

Images