Recipe recommendation method and system based on knowledge graph, device and medium
By acquiring pulse data and constructing a knowledge graph model, the problem of pulse diagnosis instruments being unable to fully acquire pulse information and insufficient adaptability of diet recommendations in existing technologies has been solved, achieving highly adaptable personalized diet recommendations.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- WUYI UNIV
- Filing Date
- 2023-04-18
- Publication Date
- 2026-06-05
AI Technical Summary
Existing pulse diagnosis devices cannot fully acquire pulse information, resulting in an inability to accurately obtain users' health parameter information. The diet recommendation methods lack adaptability and cannot provide targeted diet recommendation solutions.
By acquiring pulse data to be tested, extracting target feature parameters, and using a knowledge graph-based pulse-recipe feature relationship model, the correspondence between pulse data and recipes is pre-stored to make recipe recommendations.
It improves the adaptability and targeting of recipe recommendations, and provides personalized recipe recommendation schemes based on users' health parameters.
Smart Images
Figure CN116595186B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of recipe recommendation, and in particular to a recipe recommendation method, system, device, and medium based on knowledge graphs. Background Technology
[0002] With the rapid development of modern technology, traditional pulse diagnosis has gained worldwide attention due to its advantages such as simplicity, non-invasiveness, and painlessness, leading to the development of many different types of pulse diagnosis devices. However, current mainstream pulse diagnosis products can only measure elements such as heart rate and blood pressure, failing to comprehensively acquire pulse information, such as pulse width and pulse location, thus resulting in inaccurate acquisition of users' health parameters. Furthermore, related dietary recommendation methods simply provide corresponding dietary recommendations based on specific health parameters, lacking adaptability to different population groups; that is, the adaptability of dietary recommendations is low, thus failing to provide users with truly personalized dietary recommendations. Summary of the Invention
[0003] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention proposes a recipe recommendation method, system, device, and medium based on knowledge graphs, which can improve the adaptability and relevance of recipe recommendations.
[0004] The first aspect of this invention provides a recipe recommendation method based on a knowledge graph, comprising:
[0005] Acquire the pulse data to be tested;
[0006] Extract target feature parameters based on the pulse data to be tested;
[0007] The target feature parameters are matched with the pre-stored target pulse pattern model to determine the target pulse pattern data;
[0008] A knowledge graph-based pulse diagnosis and recipe feature relationship model is constructed, wherein the pulse diagnosis and recipe feature relationship model pre-stores the correspondence between pulse diagnosis data and recipes;
[0009] The recipe recommendation result is obtained based on the target pulse data and the pulse recipe feature relationship model.
[0010] The knowledge graph-based recipe recommendation method provided by the present invention has at least the following beneficial effects: acquiring pulse data to be tested; extracting target feature parameters from the pulse data to be tested; matching the target feature parameters with a pre-stored target pulse model to determine the target pulse data; constructing a knowledge graph-based pulse-recipe feature relationship model, which pre-stores the correspondence between pulse data and recipes; and obtaining recipe recommendation results based on the target pulse data and the pulse-recipe feature relationship model. According to the technical solution of the present invention, target pulse data can be determined by extracting target feature parameters from the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model. Since the constructed knowledge graph-based pulse-recipe feature relationship model pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendation results in the pulse-recipe feature relationship model, thereby improving the adaptability and targeting of recipe recommendations.
[0011] According to some embodiments provided by the present invention, the step of extracting target feature parameters based on the pulse data to be tested includes:
[0012] The pulse data to be tested is denoised to obtain the target feature parameters.
[0013] According to some embodiments provided by the present invention, the target feature parameters include time-domain parameters, and the recipe recommendation method further includes:
[0014] The time-domain parameters of the pulse data to be measured are processed by Fast Fourier Transform to obtain the target twelve harmonic amplitude ratio.
[0015] According to some embodiments of the present invention, the step of matching the target feature parameters with a pre-stored target pulse pattern model to determine the target pulse pattern data includes:
[0016] Subtracting the target 12th harmonic amplitude ratio from the preset 12th harmonic amplitude ratio yields the harmonic amplitude difference.
[0017] The harmonic amplitude difference is normalized by eigenvalues to obtain the error value of the harmonic amplitude difference;
[0018] The target pulse model is matched based on the error value;
[0019] The target pulse data is determined based on the target pulse model.
[0020] According to some embodiments of the present invention, the construction of a pulse diagnosis recipe feature relationship model based on knowledge graph includes:
[0021] Retrieve a list of concept entities based on a knowledge graph;
[0022] Extract entities from the concept entity list and establish a relationship graph between the entities;
[0023] Based on the relationship diagram, establish the pulse pattern diet feature relationship model.
[0024] According to some embodiments of the present invention, obtaining the recipe recommendation result based on the target pulse data and the pulse recipe feature relationship model includes:
[0025] Based on the target pulse data and the pulse diet feature relationship model, M diet recommendation schemes are generated, where M is a natural number greater than 0;
[0026] The M recommended recipes are scored and ranked to obtain a list of recommended recipes;
[0027] The first N recipe recommendations in the recipe recommendation list are taken as the recipe recommendation results, where N is a natural number greater than 0 and less than M.
[0028] According to some embodiments of the present invention, generating M recipe recommendation schemes based on the target pulse data and the pulse recipe feature relationship model includes:
[0029] The target pulse type is determined based on the target pulse data;
[0030] Based on the target pulse type, target feature information is determined in the pulse pattern diet feature relationship model;
[0031] Based on the target feature information and the preset dietary profile features, M recipe recommendation schemes are generated.
[0032] A second aspect of the present invention provides a knowledge graph-based recipe recommendation system, comprising:
[0033] The data acquisition unit is used to acquire the pulse data to be measured.
[0034] The feature extraction unit is used to extract target feature parameters based on the pulse data to be tested;
[0035] The model matching unit is used to match the target feature parameters with the pre-stored target pulse model to determine the target pulse data;
[0036] The model building unit is used to build a pulse pattern recipe feature relationship model based on knowledge graphs. The pulse pattern recipe feature relationship model pre-stores the correspondence between pulse pattern data and recipes.
[0037] The recipe recommendation unit is used to obtain recipe recommendation results based on the target pulse data and the pulse recipe feature relationship model.
[0038] The knowledge graph-based recipe recommendation system provided by the present invention has at least the following beneficial effects: acquiring pulse data to be tested; extracting target feature parameters from the pulse data to be tested; matching the target feature parameters with a pre-stored target pulse model to determine the target pulse data; constructing a knowledge graph-based pulse-recipe feature relationship model, which pre-stores the correspondence between pulse data and recipes; and obtaining recipe recommendation results based on the target pulse data and the pulse-recipe feature relationship model. According to the technical solution of the present invention, target pulse data can be determined by extracting target feature parameters from the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model. Since the constructed knowledge graph-based pulse-recipe feature relationship model pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendation results in the pulse-recipe feature relationship model, thereby improving the adaptability and targeting of recipe recommendations.
[0039] A third aspect of the present invention provides an operating apparatus, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the knowledge graph-based recipe recommendation method as described in the first aspect embodiment.
[0040] The operating apparatus provided by the embodiments of the present invention has at least the following beneficial effects: acquiring pulse data to be tested; extracting target feature parameters based on the pulse data to be tested; matching the target feature parameters with a pre-stored target pulse model to determine the target pulse data; constructing a pulse recipe feature relationship model based on a knowledge graph, wherein the pulse recipe feature relationship model pre-stores the correspondence between pulse data and recipes; and obtaining recipe recommendation results based on the target pulse data and the pulse recipe feature relationship model. According to the technical solution of the present invention, target pulse data can be determined by extracting target feature parameters from the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model; since the constructed pulse recipe feature relationship model based on a knowledge graph pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendation results in the pulse recipe feature relationship model, thereby improving the adaptability and targeting of recipe recommendations.
[0041] A fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions for performing the knowledge graph-based recipe recommendation method described in the first aspect embodiment.
[0042] The computer-readable storage medium provided by the present invention has at least the following beneficial effects: acquiring pulse data to be tested; extracting target feature parameters from the pulse data to be tested; matching the target feature parameters with a pre-stored target pulse model to determine the target pulse data; constructing a pulse recipe feature relationship model based on a knowledge graph, wherein the pulse recipe feature relationship model pre-stores the correspondence between pulse data and recipes; and obtaining recipe recommendation results based on the target pulse data and the pulse recipe feature relationship model. According to the technical solution of the present invention, target pulse data can be determined by extracting target feature parameters from the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model. Since the constructed pulse recipe feature relationship model based on a knowledge graph pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendation results in the pulse recipe feature relationship model, thereby improving the adaptability and targeting of recipe recommendations. Attached Figure Description
[0043] Additional aspects and advantages of the invention will become apparent and readily understood in conjunction with the following description of the embodiments, in which:
[0044] Figure 1 This is a flowchart of a recipe recommendation method provided in one embodiment of the present invention;
[0045] Figure 2 yes Figure 1 A flowchart illustrating the specific method of step S120;
[0046] Figure 3 This is a flowchart of a recipe recommendation method provided in another embodiment of the present invention;
[0047] Figure 4 yes Figure 1 A flowchart illustrating the specific method of step S130;
[0048] Figure 5 yes Figure 1 A flowchart illustrating the specific method of step S140;
[0049] Figure 6 yes Figure 1 A flowchart illustrating the specific method of step S150;
[0050] Figure 7 yes Figure 6 A flowchart illustrating the specific method of step S610;
[0051] Figure 8 This is a time-domain characteristic parameter diagram of the pulse to be tested provided in an embodiment of the present invention;
[0052] Figure 9 This is a schematic diagram of the structure of an operating device provided in one embodiment of the present invention. Detailed Implementation
[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0054] It is understandable that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, or the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.
[0055] Knowledge graphs, known in library and information science as knowledge domain visualization or knowledge domain mapping maps, are a series of various graphics that display the development process and structural relationships of knowledge. They use visualization techniques to describe knowledge resources and their carriers, and to mine, analyze, construct, draw, and display knowledge and the interrelationships between them. It is a modern theory that combines theories and methods from applied mathematics, computer graphics, information visualization technology, and information science with methods such as bibliometric citation analysis and co-occurrence analysis. It uses visualized graphs to vividly display the core structure, development history, cutting-edge fields, and overall knowledge architecture of a discipline, achieving the goal of multidisciplinary integration. It can provide practical and valuable references for disciplinary research.
[0056] This invention provides a knowledge graph-based recipe recommendation method, system, device, and medium, which can improve the adaptability and relevance of recipe recommendations.
[0057] The embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0058] like Figure 1 As shown, Figure 1 This is a flowchart of a knowledge graph-based recipe recommendation method provided in one embodiment of this application. The recipe recommendation method may include, but is not limited to, steps S110, S120, S130, S140, and S150.
[0059] Step S110: Obtain the pulse data to be tested;
[0060] It is understood that the pulse data to be measured refers to the pulse wave, which is the pulsation (vibration) of the heart moving along the arteries and blood flow to the surrounding area. In this embodiment, pulse data (pulse wave) can be collected through a pulse diagnosis platform. Specifically, using a pulse sensor and a stepper motor, an automatically adjustable pressure device can be built. The pulse sensor is placed on a slide, and the horizontal direction of the pressure is adjusted by a servo motor. Multiple degrees of freedom of extension and contraction can be built to simulate the human hand pressing the pulse. In addition, a visual recognition module can be used to acquire the pulse data to be measured.
[0061] It should be noted that the pulse data to be tested can be pulse data obtained in real time or pulse data obtained from a database.
[0062] Step S120: Extract target feature parameters based on the pulse data to be tested;
[0063] It is understandable that target feature parameters can include time-domain parameters and frequency-domain parameters. Time-domain analysis can be used to analyze the pulse data to obtain its time-domain parameters; frequency-domain analysis can be used to analyze the pulse data to obtain its frequency-domain parameters.
[0064] It should be noted that the target characteristic parameters include at least one of the following: pulse position, beat rate, pulse length, pulse width, pulse strength, tension, and fluency of the pulse to be measured. These target characteristic parameters allow for comprehensive and accurate acquisition of specific information about the pulse to be measured. Specifically, the depth of the pulse is called the pulse position, which is represented by the pressure applied (F); the number of pulse beats per second is called the beat rate, or heart rate (1 / T); the axial length of the pulse is called the pulse length, or pulse period (T); the radial range of the pulse is called the pulse width, or w; the intensity of the pulse is called the pulse strength, or the main wave amplitude (h1); and the degree of vasoconstriction is called tension, or diastolic area (A). d The smoothness of the pulse's arrival is called fluency, which is the area A of the systolic phase. s .
[0065] Specifically, based on the principles of time-domain waveforms and preset blood fluctuations, morphological measurements can be performed on the above target feature parameters, such as... Figure 8 As shown, the peak and trough times and values of the pulse wave within one pulse cycle can be found using differential equations in the time domain. In the figure, h1 represents the amplitude of the main wave, h2 represents the amplitude of the main wave spur, h3 represents the amplitude of the tidal wave, h4 represents the amplitude of the descending spur, h5 represents the amplitude of the dicrotic wave, k represents the slope of the main wave, T represents the pulse cycle, t1 represents the phase of the main wave, t2 represents the phase of the notch, t3 represents the phase of the main wave spur, t4 represents the phase of the descending spur, and t5 represents the phase of the dicrotic wave. d A represents the diastolic area. sThe area represents the systolic phase, and W represents 1 / 3 of the pulse width.
[0066] Step S130: Match the target feature parameters with the pre-stored target pulse model to determine the target pulse data;
[0067] It is understandable that target pulse data refers to the pulse data corresponding to the pulse to be tested in the target pulse model after the target feature parameters are matched with the pre-stored target pulse model.
[0068] Specifically, in this embodiment, the database pre-stores 28 target pulse models. These 28 target pulse models may include floating pulse, deep pulse, slow pulse, rapid pulse, slippery pulse, hesitant pulse, weak pulse, strong pulse, long pulse, short pulse, surging pulse, faint pulse, tense pulse, slow pulse, wiry pulse, hollow pulse, leathery pulse, firm pulse, soft pulse, weak pulse, scattered pulse, fine pulse, hidden pulse, arterial pulse, hurried pulse, knotted pulse, intermittent pulse, and rapid pulse. The pulse data of different target pulse models are different. By matching the target pulse feature parameters with the pre-stored 28 target pulse models, the target pulse data can be obtained.
[0069] Step S140: Construct a pulse pattern recipe feature relationship model based on knowledge graph. The pulse pattern recipe feature relationship model pre-stores the correspondence between pulse pattern data and recipes.
[0070] It is understandable that, since the pulse-based recipe feature relationship model pre-stores the correspondence between pulse data and recipes, constructing a pulse-based recipe feature relationship model based on knowledge graphs can link target pulse data with recipes, so that recipe recommendation results can be obtained subsequently based on the target pulse data and the pulse-based recipe feature relationship model.
[0071] Step S150: Obtain the recipe recommendation result based on the relationship model between the target pulse data and the pulse recipe features.
[0072] It is understandable that, since the pulse-recipe feature relationship has a pre-stored correspondence between pulse data and recipes, a corresponding recipe recommendation can be found in the pulse-recipe feature relationship model based on the target pulse data, thus obtaining the recipe recommendation result.
[0073] Understandably, according to the technical solution of the present invention, the target pulse data can be determined by extracting the target feature parameters of the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model. Since the constructed knowledge graph-based pulse recipe feature relationship model pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendations within the pulse recipe feature relationship model, thereby improving the adaptability and relevance of the recipe recommendations. Compared with related technologies, this recipe recommendation method can improve the matching similarity of the model by matching the target feature parameters of the pulse data to be tested with a pre-stored target pulse model under combined time and frequency domain analysis, thus providing users with targeted recipe recommendation solutions.
[0074] like Figure 2 As shown, in one embodiment, the recipe recommendation method is further described, and step S120 may include, but is not limited to, step S210.
[0075] Step S210: Denoise the pulse data to be tested to obtain the target feature parameters.
[0076] Understandably, the pulse data to be tested requires at least 30 cycles of data to eliminate random errors, and the pulse data to be tested needs to be transferred to memory for noise reduction processing via Direct Memory Access (DMA).
[0077] Specifically, for noisy pulse signals in the pulse data to be measured, wavelet denoising algorithms can be used to remove interference such as baseline drift and high-frequency interference. After denoising using wavelet transform threshold denoising, the main noise of the pulse signal can be suppressed, thereby accurately preserving the peaks that reflect the characteristics of the original pulse signal. For example, in one embodiment, the waveform of about 0.2 to 1.2 Hz can be preserved, and the baseline signal of 0 to 0.2 Hz can be subtracted from the pulse wave signal.
[0078] like Figure 3 As shown, in one embodiment, the recipe recommendation method is further described, the target feature parameters include time-domain parameters, and the recipe recommendation method further includes, but is not limited to, step S310.
[0079] Step S310: Perform fast Fourier transform on the time domain parameters of the pulse data to be tested to obtain the target twelve harmonic amplitude ratio.
[0080] It is understandable that by performing a fast Fourier transform on the time-domain parameters of the pulse data to be tested, the amplitude of the pulse data at different frequencies in the frequency domain can be obtained, i.e., the target twelve harmonic amplitude ratio.
[0081] like Figure 4As shown, in one embodiment, to improve the matching similarity of the model, a significance level detection can be performed during the process of matching the target feature parameters with the target pulse model to determine the target pulse data. Specifically, to further explain the recipe recommendation method, step S130 may include, but is not limited to, steps S410, S420, S430, and S440.
[0082] Step S410: Subtract the target twelfth harmonic amplitude ratio from the preset twelfth harmonic amplitude ratio to obtain the harmonic amplitude difference;
[0083] It should be noted that, in this embodiment, the preset twelfth harmonic amplitude in the database is shown in Table 1.
[0084] Table 1 Preset Twelfth Harmonic Amplitude Ratio
[0085]
[0086] Understandably, the harmonic amplitude difference can be obtained by subtracting the preset 12th harmonic amplitude ratio from the target 12th harmonic amplitude ratio in the database. The specific calculation formula is as follows:
[0087] E i (j)=[S(j)-m i (j) / m i (j)
[0088] It should be noted that E i (j) represents the harmonic amplitude difference, S(j) represents the preset twelve harmonic amplitude ratio, m i (j) represents the target twelfth harmonic amplitude ratio, where i is the sequence number of the 28 pulse signals and j is the sequence number of the twelfth harmonic.
[0089] Step S420: Perform eigenvalue normalization on the harmonic amplitude difference to obtain the error value of the harmonic amplitude difference;
[0090] Understandably, by normalizing the eigenvalues of the measured pulse image and the preset harmonic amplitude difference in the database using the mean and variance, the error value of the harmonic amplitude difference can be obtained. The specific calculation formula is as follows:
[0091]
[0092]
[0093] It should be noted that Err(i) represents the mean value of the harmonic amplitude difference, and σ(i) represents the error value.
[0094] Step S430: Match the target pulse model based on the error value;
[0095] It should be noted that matching the target pulse model using harmonic amplitude ratios will result in ratio errors, meaning the first harmonic will always be 100%. To avoid this problem, time-domain parameters can be introduced to constrain the matching of the target pulse model. The main calculation indicators are radial artery gain index rAI, blood pressure reflex index RI, notch height index RHI, rapid ejection index REI, and enhancement index AIx; the specific calculation formulas are as follows:
[0096]
[0097]
[0098]
[0099]
[0100]
[0101] In the formula, h1 represents the amplitude of the main wave, h3 represents the amplitude of the tidal wave, h4 represents the amplitude of the descending isthmus, and h5 represents the amplitude of the diurnal wave. These have been described in the above embodiments and will not be repeated here.
[0102] Understandably, by combining the 12th harmonic amplitude ratio, time-domain calculation indicators, and time-domain parameters, a parameter table can be compiled for comparison and matching. The degree of matching between group data is judged by the significance level α.
[0103] It should be noted that significance testing is a method used to detect whether there are differences between the experimental group and the control group in a scientific experiment, and whether these differences are significant. In this embodiment, when the significance level α of the target feature parameters of the pulse image data to be tested and the target pulse image model data is α < 0.05, the error rate of the matching conclusion between the two can be considered an "impossible event," that is, the target pulse image model and the target feature parameters of the pulse image to be tested are similar, and the matching degree is high.
[0104] In one embodiment, matching the target feature parameters with a pre-stored target pulse pattern model can also determine whether there is a significant difference between the target pulse pattern model and the target feature parameters. A p-value can be calculated during the left-side test by assuming a significant difference. The smaller the p-value, the stronger the reason for rejecting the null hypothesis; that is, reducing the p-value can improve the matching degree between the target feature parameters and the target pulse pattern model, and also improve the accuracy of the obtained target feature parameters. Specifically, the specific left-side test formula is as follows:
[0105] P = P(Z≤ZC|u = u0)
[0106] Where Z represents the test statistic, ZC represents the test statistic value calculated from the sample data, and the P-value represents the probability that the test statistic Z is less than the sample test statistic value ZC.
[0107] Specifically, the matching steps in this embodiment are as follows: For data from the time-domain analysis method, the normality and homogeneity of variance of the data are determined using the Shapiro-Wilk test (SW test) and the Levene test for homogeneity of variance. For variables that are normally distributed and have equal population variances, an independent samples t-test is used, expressed as mean ± standard deviation. The remaining variables are expressed using the non-parametric Mann-Whitney test. For data from the frequency-domain analysis method, the variance is used to reflect the degree of deviation between the test data and the average value of the model pulse pattern. The group with the smallest deviation is selected as the similarity group. The test group consists of 200 data points for statistical calculation. For example, in one embodiment, taking the control group as the wiry pulse from the Twenty-Eight Pulses of Traditional Chinese Medicine, the target pulse pattern model matching parameter table is shown in Table 2.
[0108] Table 2 Model Matching Table
[0109] Feature parameters Collected pulse waves (test group) String pulse model (control group) Variance / P-value <![CDATA[S0 Heart]]> 97％(100％) 100％ Variance = 3.238 <![CDATA[S1 Liver]]> 71.541(74.213％) 72.818％ Variance = 0.5654 …… …… …… …… <![CDATA[S 10 Stomach 1.187％(1.242％) 1.197％ Variance = 0.005 <![CDATA[S6 Gallbladder]]> 1.216％(1.380％) 1.325％ Variance = 0.032 <![CDATA[Amplitude of the main wave of H1]]> 558 633 P-value < 0.05 <![CDATA[Amplitude of the descending aortic notch of the H4 wave]]> 319 604 P-value < 0.05 <![CDATA[Amplitude of H5 dicrotic wave]]> 410 544 P-value = 0.049 …… …… …… …… rAI radial artery gain index 0.936±0.098 0.853±0.078 P-value = 0.045 REI Rapid Ejaculation Index 0.441 0.292 P-value < 0.05 Aix Enhancement Index -0.063±0.099 -0.147±0.078 P-value = 0.047
[0110] Understandably, by matching the target feature parameters extracted from the pulse image data with 28 pre-stored target pulse image models in the database, the variance or p-value of the comparison between each target feature parameter and the target pulse image model group can be generated. Since a p-value <0.05 is considered statistically significant in related techniques, the highest matching degree between the target feature parameters and the pre-stored target pulse image models is determined when the p-value <0.05 is the highest and the variance deviation is the smallest. Using this method to detect model matching can improve the matching degree between the target pulse image model and the target feature parameters, which is beneficial for extracting target pulse image data in subsequent steps.
[0111] Step S440: Determine the target pulse data based on the target pulse model.
[0112] Understandably, by using the above method, a target pulse model corresponding to the target feature parameters can be matched from the 28 target pulse models pre-stored in the database. This target pulse model has the highest similarity to the target feature parameters, and the target pulse data corresponding to the pulse to be tested can be determined based on this target pulse model.
[0113] like Figure 5 As shown, in one embodiment, the recipe recommendation method is further described, and step S140 may include, but is not limited to, steps S510, S520 and S530.
[0114] Step S510: Obtain a list of concept entities based on the knowledge graph;
[0115] Understandably, building a pulse feature model based on a knowledge graph requires obtaining a list of concept entities based on the knowledge graph. This list can include the concept's definition and attribute information.
[0116] For example, in one embodiment, five concepts can be defined: dish, ingredient, nutrient, ingredient type, and pulse. The specific attributes of these five concept definitions can also be described, such as the ingredients and nutrients of the dish. This can improve the nutritional similarity of the dish's characteristics to other dishes. Specifically, examples of concept definitions in the concept entity list are shown in Table 3, and examples of attribute values in the concept entity list are shown in Table 4.
[0117] Step S520: Extract entities from the concept entity list and establish a relationship graph between entities;
[0118] Understandably, by extracting entities from the list of conceptual entities and associating different entities, a relationship graph between entities can be established.
[0119] For example, in one implementation, entity extraction, by linking the relationships between entities such as "dishes," "ingredients," "nutrients," "types of ingredients," and "physiological and pathological attributes of pulse diagnosis," extracts "restrictions," "avoids," and "suitable foods" from the processed knowledge graph data, thus binding the "physiological and pathological attributes of pulse diagnosis" with "dishes," "ingredients," "nutrients," and "types of ingredients." Specifically, examples of entity extraction between the physiological and pathological attributes of pulse diagnosis and "dishes," "ingredients," "nutrients," and "types of ingredients" are shown in Table 5.
[0120] Table 3 Conceptual Entity List Conceptual Definitions
[0121]
[0122] Table 4. List of Entity Attributes in Knowledge Graph
[0123]
[0124] Table 5. List of Entity Extraction Examples
[0125]
[0126] Step S530: Establish a pulse pattern diet feature relationship model based on the relationship diagram.
[0127] It is understandable that, since the relationship graph contains the correspondence between pulse patterns and recipes, establishing a pulse pattern-recipe feature relationship model based on the relationship graph can facilitate recipe recommendation operations in subsequent steps, thereby improving the suitability of recipe recommendations.
[0128] like Figure 6 As shown, in one embodiment, the recipe recommendation method is further described, and step S150 may include, but is not limited to, steps S610, S620 and S630.
[0129] Step S610: Generate M recipe recommendations based on the target pulse data and the pulse recipe feature relationship model, where M is a natural number greater than 0;
[0130] It is understandable that the target pulse data can include multiple pieces of information, and one pulse pattern can correspond to multiple recipe schemes based on the pulse pattern recipe feature relationship model. Therefore, multiple recipe recommendation schemes can be generated based on the target pulse data and the pulse pattern recipe feature relationship model.
[0131] Step S620: Rate and sort the M recipe recommendations to obtain a recipe recommendation list;
[0132] Understandably, a list of recommended recipes can be obtained by rating and ranking multiple recipe recommendations. Specifically, the rating formula for the recipe recommendations is as follows:
[0133]
[0134] Where Gui represents the comprehensive score of the dietary therapy recommendation plan, Sim(i,j) represents the final similarity between dish i and dish j, A(i,z) represents the z most similar recipes to recipe i, B(u) represents the most needed recipe in pulse diagnosis dietary therapy, and A(i,z)∩B(u) represents the intersection of the two recipe sets. It should be noted that the specific calculation methods of Sim(i,j) and Rmn will be explained in steps S710, S720 and S730.
[0135] Step S630: Take the first N recipe recommendations from the recipe recommendation list as the recipe recommendation result, where N is a natural number greater than 0 and less than M.
[0136] Understandably, sorting the calculated comprehensive score Gui yields a list of recommended recipes; selecting the top N recipe recommendations from this list improves the suitability and relevance of recipes to different groups of people.
[0137] like Figure 7 As shown, in one embodiment, the recipe recommendation method is further described, and step S610 may include, but is not limited to, steps S710, S720 and S730.
[0138] Step S710: Determine the target pulse type based on the target pulse data;
[0139] Step S720: Determine the target feature information in the pulse pattern diet feature relationship model according to the target pulse pattern type;
[0140] Step S730: Generate M recipe recommendation schemes based on target feature information and preset dietary profile features.
[0141] It is understandable that target pulse data refers to the pulse data corresponding to the pulse to be tested in the target pulse model, which may include the target pulse type. Based on the target pulse type, the target feature information can be determined in the pulse diet feature relationship model.
[0142] It should be noted that target feature information can reflect the user's health parameter information. For example, taking the wiry pulse among the twenty-eight pulse types as an example, when the target pulse type is determined to be the wiry pulse, the feature information of the wiry pulse in the pulse pattern diet feature relationship model is liver qi stagnation, liver and gallbladder damp heat, etc., which can reflect the user's health parameter information. Thus, based on the health parameter information, it can be determined that the user is in a state of symptoms such as qi stagnation, blood stasis, and pain.
[0143] It should be noted that the preset dietary profile features can be obtained from a database or input based on user preferences. Combining dietary profile features, such as dietary preferences, cooking preferences, human nutritional and health characteristics, BMI, allergy history, and cost sensitivity to dishes, the user's preferred dishes can be weighted according to taste, price, and cooking method. Based on the target feature information and preset dietary profile features, multiple personalized combination recipe recommendations can be generated for the user to choose from, enriching the diversity of recipe recommendations.
[0144] Specifically, the process of generating M recipe recommendation schemes based on target feature information and preset dietary profile features is as follows:
[0145] (1) Weighting of unique dishes based on dietary profile characteristics. In one embodiment, a string search algorithm (Knuth-Morris-Pratt, KMP) is used for matching. The daily nutrient intake of the diet of the patient to be tested is weighted highest (a), the user's preference for dishes based on dietary profile characteristics is weighted second (b), the user's price sensitivity based on dietary profile characteristics is weighted third (c), and the patient's own contraindications are weighted negatively (d).
[0146] Let U = {U1, U2, ..., Um} be the user's preference set, with m types of preferences; let W = {W1, W2, ..., W3} be the number of dishes, with n dishes; the formula for calculating the preference-dish rating matrix Rmn is as follows.
[0147]
[0148] It should be noted that Rij represents the combined rating of user preference (UI) and dish (Wj), and the system can make recommendations based on the rating.
[0149] (2) Users describe a feature differently, so it is necessary to calculate dish similarity and semantic similarity. The similarity calculation can ensure the substitutability of dishes with the same effect but different ingredients and cooking methods.
[0150] Specifically, the cosine similarity Sim(i,j) can be used to calculate the similarity between dish i and dish j, where S... i and S j S represents the content of the ingredients in dish i and dish j, respectively. ki S represents the content of ingredient k in dish i. kj Let be the content of ingredient k in dish j. The cosine similarity (Sim) of the dishes. cat The formula for calculating (i,j) is shown below.
[0151]
[0152] Semantically, listing the relationships between various entity attributes, the TranHR model can be used to handle 1-N type relationships, i.e., a recipe for a dish corresponds to multiple target pulse data, and an ingredient corresponds to multiple dishes. Its functional representation is Sim. cf (i,j). Simultaneous cosine similarity of the dishes. cat (i,j) is the function Sim obtained by combining the Tran HR model under the knowledge graph. cf The similarity (i,j) is fused using a linear weighted average, where α is the weight coefficient and 0 < α < 1, to obtain the final similarity between dish i and dish j. The specific formula for calculating the final similarity Sim(i,j) is shown below:
[0153] Sim(i,j)=α*Sim cat (i,j)+(1-α)Sim cf (i,j)
[0154] A second aspect of the present invention provides a knowledge graph-based recipe recommendation system, comprising:
[0155] The data acquisition unit is used to acquire the pulse data to be measured.
[0156] The feature extraction unit is used to extract target feature parameters based on the pulse data to be tested.
[0157] The model matching unit is used to match the target feature parameters with the pre-stored target pulse model to determine the target pulse data.
[0158] The model building unit is used to build a pulse pattern recipe feature relationship model based on knowledge graphs. The pulse pattern recipe feature relationship model pre-stores the correspondence between pulse pattern data and recipes.
[0159] The recipe recommendation unit is used to obtain recipe recommendation results based on the target pulse data and the relationship model between pulse and recipe features.
[0160] Understandably, according to the technical solution of the present invention, the following steps are taken: acquiring pulse data to be tested; extracting target feature parameters from the pulse data to be tested; matching the target feature parameters with a pre-stored target pulse model to determine the target pulse data; constructing a pulse recipe feature relationship model based on a knowledge graph, wherein the pulse recipe feature relationship model pre-stores the correspondence between pulse data and recipes; and obtaining recipe recommendation results based on the target pulse data and the pulse recipe feature relationship model. According to the technical solution of the present invention, target pulse data can be determined by extracting target feature parameters from the pulse data to be tested and matching the target feature parameters with a pre-stored target pulse model. Since the constructed pulse recipe feature relationship model based on a knowledge graph pre-stores the correspondence between pulse data and recipes, the target pulse data can obtain corresponding recipe recommendation results in the pulse recipe feature relationship model, thereby improving the adaptability and targeting of recipe recommendations.
[0161] like Figure 9 As shown, a third aspect embodiment of the present invention provides an operating device 800, including: a memory 810, a processor 820, and a computer program stored in the memory 810 and executable on the processor 820. When the processor 820 executes the computer program, it implements the knowledge graph-based recipe recommendation method as described in the first aspect embodiment.
[0162] The processor 820 and memory 810 can be connected via a bus or other means.
[0163] The non-transient software program and instructions required to implement the knowledge graph-based recipe recommendation method of the above embodiments are stored in memory 810. When executed by processor 820, the knowledge graph-based recipe recommendation method of the above embodiments is executed, for example, the method described above is executed. Figure 1 Method steps S110 to S150 in the middle Figure 2 Method steps S210, Figure 3 Method steps S310, Figure 4 Method steps S410 to S440 in the middle Figure 5 Method steps S510 to S530, Figure 6 Method steps S610 to S630 in the middle Figure 7 Method steps S710 to S730.
[0164] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.
[0165] A fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions that are executed by a processor or controller, for example, by a processor 820 in the above-described operating device 800 embodiment, causing the processor 820 to execute the knowledge graph-based recipe recommendation method in the above-described embodiments, for example, to perform the above-described... Figure 1 Method steps S110 to S150 in the middle Figure 2 Method steps S210, Figure 3 Method steps S310, Figure 4 Method steps S410 to S440 in the middle Figure 5 Method steps S510 to S530, Figure 6 Method steps S610 to S630 in the middle Figure 7 Method steps S710 to S730.
[0166] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, and the base station system, can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.
[0167] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0168] The above is a detailed description of the preferred embodiments of this application. However, this application is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.
Claims
1. A recipe recommendation method based on knowledge graphs, characterized in that, include: Acquire the pulse data to be tested; Extract target feature parameters based on the pulse data to be tested; The target feature parameters are matched with a pre-stored target pulse pattern model to determine the target pulse pattern data; the target feature parameters include time-domain parameters, which are used to constrain the matching of the target pulse pattern model. Where h1 represents the amplitude of the main wave, h3 represents the amplitude of the tidal wave, h4 represents the amplitude of the descending isthmus, h5 represents the amplitude of the dicrotic wave, rAI represents the radial artery gain index, RI represents the blood pressure reflex index, RHI represents the notch height index, REI represents the rapid ejection index, and AIx represents the enhancement index. The time-domain parameters of the pulse data to be measured are processed by Fast Fourier Transform to obtain the target twelve harmonic amplitude ratio; The step of matching the target feature parameters with a pre-stored target pulse pattern model to determine the target pulse pattern data includes: subtracting the target 12th harmonic amplitude ratio from the preset 12th harmonic amplitude ratio to obtain a harmonic amplitude difference; performing eigenvalue normalization processing on the harmonic amplitude difference to obtain an error value of the harmonic amplitude difference; matching the target pulse pattern model according to the error value; and determining the target pulse pattern data according to the target pulse pattern model. A knowledge graph-based pulse diagnosis and recipe feature relationship model is constructed, wherein the pulse diagnosis and recipe feature relationship model pre-stores the correspondence between pulse diagnosis data and recipes; The recipe recommendation result is obtained based on the target pulse data and the pulse recipe feature relationship model.
2. The recipe recommendation method according to claim 1, characterized in that, The step of extracting target feature parameters based on the pulse data to be tested includes: The pulse data to be tested is denoised to obtain the target feature parameters.
3. The recipe recommendation method according to claim 1, characterized in that, The construction of a pulse diagnosis recipe feature relationship model based on knowledge graphs includes: Retrieve a list of concept entities based on a knowledge graph; Extract entities from the concept entity list and establish a relationship graph between the entities; Based on the relationship diagram, establish the pulse pattern diet feature relationship model.
4. The recipe recommendation method according to claim 1, characterized in that, The step of obtaining the recipe recommendation result based on the target pulse data and the pulse recipe feature relationship model includes: Based on the target pulse data and the pulse diet feature relationship model, M diet recommendation schemes are generated, where M is a natural number greater than 0; The M recommended recipes are scored and ranked to obtain a list of recommended recipes; The first N recipe recommendations in the recipe recommendation list are taken as the recipe recommendation results, where N is a natural number greater than 0 and less than M.
5. The recipe recommendation method according to claim 4, characterized in that, The process of generating M recipe recommendation schemes based on the target pulse data and the pulse recipe feature relationship model includes: The target pulse type is determined based on the target pulse data; Based on the target pulse type, target feature information is determined in the pulse pattern diet feature relationship model; Based on the target feature information and the preset dietary profile features, M recipe recommendation schemes are generated.
6. A recipe recommendation system based on knowledge graphs, characterized in that, include: The data acquisition unit is used to acquire the pulse data to be measured. The feature extraction unit is used to extract target feature parameters based on the pulse data to be tested; A model matching unit is used to match the target feature parameters with a pre-stored target pulse pattern model to determine the target pulse pattern data; the target feature parameters include time-domain parameters, which constrain the matching of the target pulse pattern model. Where h1 represents the amplitude of the main wave, h3 represents the amplitude of the tidal wave, h4 represents the amplitude of the descending isthmus, h5 represents the amplitude of the dicrotic wave, rAI represents the radial artery gain index, RI represents the blood pressure reflex index, RHI represents the notch height index, REI represents the rapid ejection index, and AIx represents the enhancement index. The time-domain parameters of the pulse data to be measured are processed by Fast Fourier Transform to obtain the target twelve harmonic amplitude ratio; The step of matching the target feature parameters with a pre-stored target pulse pattern model to determine the target pulse pattern data includes: subtracting the target 12th harmonic amplitude ratio from the preset 12th harmonic amplitude ratio to obtain a harmonic amplitude difference; performing eigenvalue normalization processing on the harmonic amplitude difference to obtain an error value of the harmonic amplitude difference; matching the target pulse pattern model according to the error value; and determining the target pulse pattern data according to the target pulse pattern model. The model building unit is used to build a pulse pattern recipe feature relationship model based on knowledge graphs. The pulse pattern recipe feature relationship model pre-stores the correspondence between pulse pattern data and recipes. The recipe recommendation unit is used to obtain recipe recommendation results based on the target pulse data and the pulse recipe feature relationship model.
7. An operating device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, it implements the knowledge graph-based recipe recommendation method as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for performing the knowledge graph-based recipe recommendation method according to any one of claims 1 to 5.