A method for evaluating the heat determination of a subject and an electronic device

Abstract

The application discloses a kind of methods for determining the heat of evaluation object and electronic equipment, it is related to data processing field, by obtaining the evaluation parameter of evaluation object under the dimension of multi-source heterogeneous data, and type division is carried out based on historical data completeness, and different processing strategies are used for evaluation object of different data completeness.For the evaluation object of data completeness, the dynamic weight associated with the region is introduced for weighted fusion processing, so that the heat determination process can adapt to the data feature difference of different regions;For the evaluation object of data imperfection, the heat level is determined based on the review data statistics of similar complete evaluation object, which solves the evaluation problem in the data missing scenario.Finally, the initial heat level is calibrated by correction processing, which improves the accuracy and reliability of the heat determination result.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for determining the heat of an evaluation object and an electronic device. Background Technology

[0002] In the field of data analysis and popularity assessment, evaluation methods typically rely on statistical analysis of data from a single data source or limited dimensions. For example, in the scenario of service venue popularity assessment, historical transaction data or user access data are used to rank or classify the evaluation objects through statistical models. These methods are limited by the singularity of the data source and struggle to comprehensively reflect the overall performance level of the evaluation objects.

[0003] Furthermore, accurately quantifying the contribution of different data sources to the analysis results under different scenarios during the fusion and processing of multi-source heterogeneous data remains a significant technical challenge. Simultaneously, obtaining reliable data analysis results for evaluation subjects with scarce historical data is also an unresolved issue. These factors lead to biases in the data processing results for the evaluation subjects, making it difficult to meet the accuracy and reliability requirements of data analysis in practical application scenarios. Summary of the Invention

[0004] In view of the above problems, this application provides a method for determining the popularity of an evaluation object, so as to improve the accuracy and adaptability of the popularity determination results. The specific solution is as follows:

[0005] The first aspect of this application provides a method for determining the popularity of an evaluation object, including:

[0006] Obtain multi-source heterogeneous data of the evaluation object, wherein the multi-source heterogeneous data includes at least production data, traffic data, review data, and exposure data;

[0007] Based on the multi-source heterogeneous data, the evaluation parameters of the evaluation object in each data dimension are determined;

[0008] Based on the completeness of the historical data of the evaluation object, the target type of the evaluation object is determined. The target type includes a first type and a second type. The first type indicates that the number of historical data records of the evaluation object in each data dimension has reached the target number range. The second type indicates that the number of historical data records of the evaluation object in at least one data dimension has not reached the target number range.

[0009] In response to the target type of the evaluation object being the first type, the evaluation parameters are weighted and fused based on dynamic weights to obtain the initial popularity level of the evaluation object. The dynamic weights represent the contribution of evaluation parameters of different data dimensions in the popularity determination process, and the value of the dynamic weights is related to the region to which the evaluation object belongs.

[0010] In response to the target type of the evaluation object being the second type, the initial popularity level of the evaluation object is determined based on the statistical data of the review data of each reference evaluation object belonging to the first type;

[0011] The initial heat level is corrected to obtain the target heat level of the evaluation object.

[0012] In one possible implementation, the step of weighting and fusing the evaluation parameters based on dynamic weights to obtain the initial popularity level of the evaluation object includes:

[0013] Obtain the region identifier of the area to which the evaluation object belongs;

[0014] Based on the region identifier, the weight values ​​of each data dimension corresponding to the region identifier are read from a preset weight configuration table, wherein the weight values ​​of each data dimension corresponding to different region identifiers are at least partially different.

[0015] Based on the weight values ​​of each data dimension, the evaluation parameters are weighted and summed to obtain the heat fusion value;

[0016] The initial popularity level of the evaluation object is determined based on the comparison result between the popularity fusion value and the preset level classification threshold.

[0017] In one possible implementation, the weight values ​​in the weight configuration table are dynamically adjusted based on the analysis results of the influence of each data dimension on the heat determination result in historical data.

[0018] In one possible implementation, the method further includes:

[0019] Based on the values ​​of the evaluation parameters in each data dimension, determine the ranking of the evaluation object in each data dimension within its region;

[0020] The evaluation parameters are adjusted based on the ranking to obtain the adjusted evaluation parameters.

[0021] In one possible implementation, determining the initial popularity level of the evaluation object based on the statistical data of each reference evaluation object belonging to the first type includes:

[0022] Obtain the point ratings for each reference assessment object belonging to the first type;

[0023] Based on the point ratings of each reference evaluation object, the mode, high popularity threshold, and low popularity threshold are determined. The mode is the most frequently occurring value among the point ratings of each reference evaluation object. The high popularity threshold is the lower limit of the point rating of reference evaluation objects belonging to the first type and having a high popularity level. The low popularity threshold is the upper limit of the point rating of reference evaluation objects belonging to the first type and having a low popularity level.

[0024] Obtain the point rating value of the evaluation object;

[0025] The initial popularity level of the evaluated object is determined based on the comparison results between the point score of the evaluated object and the mode, high popularity threshold and low popularity threshold.

[0026] In one possible implementation, determining the initial popularity level of the evaluation object based on a comparison of its point score with the mode, high popularity threshold, and low popularity threshold includes:

[0027] If the rating of the evaluated object is equal to the mode, the initial popularity level of the evaluated object is determined to be the popularity level corresponding to the mode;

[0028] If the score of the evaluated object is not less than the high popularity threshold, the initial popularity level of the evaluated object is determined to be a high popularity level;

[0029] If the point score of the evaluated object is not greater than the low popularity threshold, the initial popularity level of the evaluated object is determined to be low popularity level;

[0030] If the rating of the evaluated object is between the low popularity threshold and the high popularity threshold and is not equal to the mode, the initial popularity level of the evaluated object is determined to be the intermediate popularity level.

[0031] In one possible implementation, the step of correcting the initial popularity level to obtain the target popularity level of the evaluated object includes:

[0032] Obtain attribute data of the assessment object, wherein the attribute data includes at least one of the following: area identifier, location type, and service capacity data;

[0033] The attribute data is matched with the condition items in the preset constraint condition library. The constraint condition library contains at least one constraint condition, and each constraint condition includes a condition item and a corresponding correction operation.

[0034] If a match is successful, the initial heat level is adjusted according to the correction operation corresponding to the matched target constraint to obtain the target heat level.

[0035] In one possible implementation, the constraints in the constraint library include at least one of the following:

[0036] If the area is identified as a first preset area, the venue type is a preset type, and the service capacity data is lower than the first capacity threshold, the initial popularity level will be adjusted to a low popularity level.

[0037] If the service capacity data is higher than the second capacity threshold and the initial popularity level is the intermediate popularity level, the initial popularity level will be adjusted to the high popularity level.

[0038] If the service capacity data is lower than the third capacity threshold and the initial popularity level is high popularity level, then the initial popularity level will be adjusted to the intermediate popularity level.

[0039] One possible implementation also includes:

[0040] The target popularity level is output to the downstream system so that the downstream system can perform corresponding data processing operations on the evaluation object according to the target popularity level.

[0041] A second aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0042] The memory is used to store computer programs;

[0043] The processor is used to execute the computer program so that the electronic device can implement the method for determining the heat of the evaluation object in the first aspect or any implementation thereof.

[0044] By employing the aforementioned technical solution, this application provides a method and electronic device for determining the popularity of an evaluation object. This method acquires evaluation parameters of the evaluation object across multiple heterogeneous data sources and categorizes the evaluation object based on historical data completeness. Differentiated processing strategies are then applied to evaluation objects with varying levels of data completeness. For evaluation objects with complete data, dynamic weights associated with specific regions are introduced for weighted fusion processing, enabling the popularity determination process to adapt to differences in data characteristics across different regions. For evaluation objects with incomplete data, popularity levels are estimated based on statistical data from reviewers of similar complete evaluation objects, addressing the evaluation challenge in scenarios with missing data. Finally, a correction process calibrates the initial popularity level, improving the accuracy and reliability of the popularity determination results. This solves technical problems such as difficulties in fusing multi-source heterogeneous data, poor regional feature adaptability, and missing scenario evaluations, achieving full coverage of different types of evaluation objects and improving the applicability and robustness of the popularity determination method. Attached Figure Description

[0045] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0046] Figure 1 A flowchart illustrating a method for determining the heat of an evaluation object, provided in an embodiment of this application;

[0047] Figure 2 A flowchart illustrating the heat stratification process for an evaluation object, provided in an embodiment of this application;

[0048] Figure 3 A flowchart of the evaluation object heat stratification and data fusion processing provided in this application embodiment;

[0049] Figure 4 This is a diagram illustrating a hotel popularity application scenario and output result architecture provided in an embodiment of this application. Detailed Implementation

[0050] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0051] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0052] The terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of units is not necessarily limited to those units, but may include other units not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0053] This application discloses a method and electronic device for determining the popularity of an evaluation object, applicable to scenarios requiring popularity analysis such as hotels, restaurants, and tourist attractions. The method first acquires multi-source heterogeneous data of the evaluation object, including at least production data, traffic data, review data, and exposure data, and determines the evaluation parameters of the evaluation object in each data dimension based on this data. Then, based on the completeness of the historical data of the evaluation object, it is divided into a first type or a second type: the first type indicates that the number of historical data records for each data dimension of the evaluation object reaches the target range, while the second type indicates that the number of historical data records in at least one data dimension does not reach the target range. For the first type of evaluation object, a weighted fusion process is performed on the evaluation parameters using dynamic weights associated with the region to obtain an initial popularity level, where the dynamic weights represent the contribution of different data dimensions in the popularity determination process; for the second type of evaluation object, its initial popularity level is determined based on the statistical data of review data of each reference evaluation object belonging to the first type. Finally, the initial popularity level is corrected to obtain the target popularity level of the evaluation object. This application achieves differentiated processing for different data completeness assessment objects through the above data processing method, effectively solving technical problems such as difficulty in fusion of multi-source heterogeneous data, poor adaptability of regional features, and lack of assessment in cold start scenarios, thereby improving the accuracy and applicability of the heat determination results.

[0054] For ease of description, hotels are primarily used as an example for evaluation in this application embodiment. It is understood that the evaluation object described in this application is not limited to hotels, but can also be restaurants, scenic spots, guesthouses, and other service venues requiring popularity evaluation. Correspondingly, the popularity determination method for the evaluation object provided in this application embodiment can be executed by electronic devices, including but not limited to servers, computer equipment, or distributed processing systems.

[0055] See Figure 1 The illustration shows a flowchart of a method for determining the popularity of an evaluation object according to an embodiment of this application. The method may include the following steps:

[0056] S101. Obtain multi-source heterogeneous data of the evaluation object.

[0057] The evaluation object represents the object whose popularity level needs to be determined in the current popularity evaluation scenario, such as hotels, shopping malls, and restaurants. Multi-source heterogeneous data refers to various types of data originating from different channels and differing in data structure and format. In the embodiments of this application, multi-source heterogeneous data includes at least output data, traffic data, review data, and exposure data. Output data represents the quantitative data of the actual operating results of the evaluation object. For example, when the evaluation object is a hotel, data such as non-cancelled room nights, number of completed orders, and transaction amount are extracted for the hotel in a preset period (such as the most recent six months) and a historical benchmark period (such as a representative year). If the evaluation object is a scenic spot or shop, output data can be replaced with actual conversion data such as the number of visitors and the number of transactions. Traffic data represents the market attention of the evaluation object. For example, when the evaluation object is a hotel, the unique visitor (UV) data of the hotel in the corresponding period is aggregated from the user behavior logs of various platforms within the group, and the rate of change compared with the same period in history can be further calculated. Traffic data is used to reflect the hotel's user attention within the platform.

[0058] Review data represents the market reputation of the evaluated entity. This can be obtained from mainstream online travel platforms, review websites, social media, and other channels, including the number of user reviews, ratings, and sentiment analysis over the past year and historical periods. Exposure data represents the market awareness of the evaluated entity. This can be collected from global search engines, local vertical travel comparison platforms, social media, and advertising channels, including data on exposure, impressions, and brand mentions.

[0059] Furthermore, to ensure the accuracy of data processing, after obtaining the aforementioned multi-source heterogeneous data, data preprocessing can be performed, such as preliminary deduplication, noise reduction, and missing value completion, to ensure data validity. This step achieves comprehensive integration of internal and external data through multi-source data acquisition, providing an accurate and comprehensive data foundation for subsequent popularity determination.

[0060] S102. Based on multi-source heterogeneous data, determine the evaluation parameters of the evaluation object in each data dimension.

[0061] Among them, the evaluation parameters refer to the comparable quantitative indicators obtained after standardizing and normalizing the multi-source heterogeneous data of each dimension. They are the basis for subsequent heat calculation and eliminate the differences in the dimensions of different data dimensions and regional market differences.

[0062] Specifically, data cleaning and normalization can be performed on multi-source heterogeneous data. Data cleaning includes handling outliers and missing values ​​in the original data corresponding to the multi-source heterogeneous data. For example, duplicate reviews with obvious anomalies in the review data are deduplicated; abnormal peaks caused by web crawler attacks in the exposure data are removed; and missing indicators can be filled with the average value of hotels of the same type in the same region. To eliminate the influence of different indicator units, the data is normalized to map the original values ​​to a uniform numerical range (e.g., 0-100 points). Taking production data as an example, the Min-Max normalization method can be used: Production evaluation parameter = [(Hotel room nights - Minimum hotel room nights in the country) / (Maximum hotel room nights in the country - Minimum hotel room nights in the country)] × 100.

[0063] For traffic data, review data, and exposure data, a similar method is used to calculate their respective evaluation parameters. For review data, the number of reviews and the rating score can be considered together. For example, the weighted score can be calculated as follows: Weighted Score = Review Quantity Weight × Review Quantity Normalized Value + Rating Weight × Rating Normalized Value.

[0064] This step eliminates the differences in dimensions, regional market differences, and platform data among multi-source heterogeneous data, making the data of different evaluation objects and different regions comparable across dimensions. This provides a unified quantitative basis for subsequent weighted fusion and benchmarking evaluation, and avoids calculation deviations caused by different data formats.

[0065] S103. Determine the target type of the assessment object based on the completeness of its historical data.

[0066] Historical data completeness refers to whether the quantity and time span of historical data records for the evaluated object across the four data dimensions of output, traffic, reviews, and exposure meet the preset evaluation requirements. The target quantity range refers to the minimum threshold quantity of historical data records for each data dimension, pre-set according to the business scenario (e.g., at least 3 months of continuous records for output data, at least 50 valid reviews for review data), and is the main basis for classifying the evaluation object type. Type 1 indicates that the evaluation object's historical data record quantity for each data dimension meets the target quantity range; that is, Type 1 is a data-complete evaluation object. This could be a core hotel or core merchant with abundant historical operational data, allowing for accurate popularity assessment through multi-dimensional weighted fusion. Type 2 indicates that the evaluation object has at least one data dimension where the number of historical data records does not meet the target quantity range; that is, Type 2 is a data-deficient evaluation object. This could be a non-core hotel or a cold-start hotel with at least one dimension lacking data or having insufficient records (e.g., newly launched homestays with no output data or only 10 reviews), making effective assessment impossible through weighted fusion.

[0067] This step enables refined stratification of the evaluation objects, providing a basis for subsequent differentiated popularity evaluation strategies, solving the problem of traditional methods failing to evaluate cold-start objects, and achieving full coverage of the evaluation objects.

[0068] S104. In response to the target type of the evaluation object being the first type, the evaluation parameters are weighted and fused based on dynamic weights to obtain the initial popularity level of the evaluation object.

[0069] Dynamic weights characterize the contribution of evaluation parameters from different data dimensions to the heat determination process, and the values ​​of dynamic weights are related to the region to which the evaluation object belongs. Weighted fusion processing refers to multiplying the evaluation parameters of each dimension with their corresponding dynamic weights and then performing a fusion calculation to obtain a comprehensive quantitative value of heat. The initial heat level refers to the heat level before it has been corrected by business rules, and is usually divided into three levels: high heat level, medium heat level, and low heat level.

[0070] Specifically, the dynamic weighting configuration for the evaluated object can be determined based on its region (e.g., different countries or different cities within the same country). Output and reviews typically have primary weights, while traffic and exposure usually have secondary weights. The weighting ratio varies across regions (e.g., region A has higher weighting for local platform traffic, while region B has higher weighting for global search engine exposure). The evaluation parameters for each data dimension are multiplied by their corresponding dynamic weights to obtain a weighted score for each dimension. The weighted scores for each dimension are then summed to obtain the object's popularity fusion value. Pre-set thresholds for popularity fusion value classifications (e.g., high popularity ≥ 8 points, medium popularity 5-8 points, low popularity < 5 points). Based on the threshold range where the popularity fusion value falls, the initial popularity level of the evaluated object is determined.

[0071] In this step, the weights are adaptively adjusted by region, which solves the problem of localized market evaluation distortion caused by static weights in traditional methods; and the weighted fusion realizes the comprehensive consideration of multi-dimensional evaluation parameters, which improves the accuracy of the popularity evaluation of the first type of evaluation object.

[0072] S105. In response to the target type of the evaluation object being the second type, the initial popularity level of the evaluation object is determined based on the statistical data of the review data of each reference evaluation object belonging to the first type.

[0073] The reference evaluation object refers to the first-type evaluation object belonging to the same region and category as the second-type evaluation object. For example, for a second-type homestay in region A, the reference evaluation object is a first-type homestay or a mid-to-high-end hotel in region A. Review data statistics refer to the characteristic values ​​obtained from the statistical analysis of the review data of the reference evaluation object, such as the mode of the rating scores and the rating score thresholds for each popularity level. These are the core benchmarks for the popularity comparison evaluation of the second-type evaluation object.

[0074] Specifically, all Type 1 evaluation objects in the same region and category as the Type 2 evaluation objects can be selected as a reference evaluation object set. The point ratings (core review data statistics) of all objects in the reference evaluation object set are extracted, and statistical analysis is performed on these point ratings to obtain review data statistics (such as mode, high popularity threshold, low popularity threshold). Valid point ratings for the Type 2 evaluation objects are extracted (even if the number is very small, such as the average score of 10 reviews). The point ratings of the Type 2 evaluation objects are compared with the review data statistics of the reference evaluation objects, and their initial popularity level is determined based on the comparison results.

[0075] This step enables effective evaluation of data-scarce assessment targets, overcoming the limitations of traditional methods that rely on historical data. By establishing market benchmarks with reference targets in the same region and product category, the rationality and regional adaptability of the popularity levels of the second type of assessment targets are achieved.

[0076] S106. Correct the initial heat level to obtain the target heat level of the evaluation object.

[0077] The correction process refers to the calibration and adjustment of the initial popularity level based on the actual attribute data of the evaluated object and preset business constraints. Attribute data of the evaluated object (such as region identifier, venue type, service capacity data, and business category) can be extracted. A business constraint library is pre-established, containing constraints that match the actual operation strategy (such as special rules for regional homestays, service capacity threshold rules, etc.), with each constraint corresponding to a specific correction operation. The attribute data of the evaluated object is matched against the conditions in the constraint library to determine whether a correction operation is triggered. If a correction operation is triggered, the initial popularity level is adjusted according to the corresponding correction operation; if not triggered, the initial popularity level is the target popularity level. For example, taking a second-type homestay in region A as an example, its initial popularity level is an intermediate popularity level. Its attribute data is extracted as follows: Region identifier: Region A, Venue type: Homestay, Service capacity data = 45 rooms. The preset constraint library includes "For homestays in area A with ≤50 rooms, adjust the initial popularity level to a low popularity level"; when the homestay matches this constraint, a correction operation is triggered, adjusting the initial intermediate popularity level to a low popularity level, i.e., the target popularity level.

[0078] By correcting the initial heat level, the mismatch between the data output and the preset constraints was resolved, and the matching degree between the heat determination result and the actual attributes of the evaluation object was improved.

[0079] This application provides a method for determining the popularity of an evaluation object. It acquires evaluation parameters of the object across multiple heterogeneous data sources and categorizes the object based on historical data completeness. Different processing strategies are employed for evaluation objects with varying data completeness. For objects with complete data, dynamic weights associated with specific regions are introduced for weighted fusion processing, enabling the popularity determination process to adapt to differences in data characteristics across different regions. For objects with incomplete data, popularity levels are estimated based on statistical data from similar complete evaluation objects, addressing the evaluation challenge in scenarios with missing data. Finally, a correction process calibrates the initial popularity level, improving the accuracy and reliability of the popularity determination results. This solves technical problems such as difficulties in multi-source heterogeneous data fusion, poor regional feature adaptability, and missing scenario evaluations, achieving full coverage of different types of evaluation objects and improving the applicability and robustness of the popularity determination method.

[0080] In one possible implementation, the initial popularity level of the evaluation object is obtained by weighting the evaluation parameters based on dynamic weights, including: obtaining the regional identifier of the region to which the evaluation object belongs; reading the weight values ​​of each data dimension corresponding to the regional identifier from a preset weight configuration table, wherein the weight values ​​of each data dimension corresponding to different regional identifiers are at least partially different; weighting and summing the evaluation parameters according to the weight values ​​of each data dimension to obtain a popularity fusion value; and determining the initial popularity level of the evaluation object based on the comparison result between the popularity fusion value and the preset level division threshold.

[0081] A regional identifier is a code or character used to uniquely identify the region to which an evaluation object belongs. It can be granularized by continent, country, or city (e.g., Asia Pacific = AP, South Korea = KR, Seoul = KR-SJ, United States = US, New York = US-NY), and serves as an index for retrieving dynamic weights. Specifically, information about the region to which the evaluation object belongs can be extracted from the object's basic information database, and the corresponding regional identifier can be matched according to preset regional granularity rules. If the evaluation object's regional information is multi-level (e.g., city-country-continent), the core regional identifier can be set at the city or country level to achieve refined weight configuration.

[0082] A weighting configuration table is a two-dimensional table pre-built based on the market characteristics, user channel preferences, and the influence analysis results of data dimensions for each region. For example, the row dimension in the constructed weighting configuration table is the region identifier, the column dimension is the data dimension (output, traffic, reviews, exposure), and the cell represents the weight value of that data dimension for the corresponding region (which can be expressed as a percentage or decimal). For instance, if the evaluation object is a type 1 hotel in region A, with a region identifier of A00, the weight values ​​read from the weighting configuration table would be: output 40%, reviews 30%, traffic 20%, and exposure 10%. As another example, if the evaluation object is a type 1 hotel in region B, with a region identifier of B01, the corresponding weight values ​​read from the weighting configuration table could be: output 35%, reviews 25%, traffic 15%, and exposure 25%.

[0083] Then, the relevant evaluation parameters are weighted and summed according to the weight values ​​of each data dimension to obtain the heat fusion value. This heat fusion value refers to the quantitative value that comprehensively reflects the overall heat of the evaluated object. For example, the heat fusion value can be calculated according to the following weighted summation formula: Heat Fusion Value = Σ (Evaluation Parameters of Each Data Dimension × Corresponding Weight Value), where the weight values ​​are used in the calculation in decimal form (e.g., 40% = 0.4) to ensure the accuracy of the calculation results; if there are multiple sub-indicators for a certain data dimension, the fusion evaluation parameters of the sub-indicators must be calculated first before participating in the weighted summation.

[0084] The initial popularity level of the evaluated object is determined by comparing the popularity fusion value with the preset level classification threshold. The level classification threshold refers to a range of popularity fusion values ​​pre-set based on business scenarios and market data distribution. This range is used to convert the quantitative popularity fusion value into a qualitative popularity level, typically divided into high, medium, and low levels. The threshold can be flexibly adjusted according to the characteristics of the regional market. For example, for core hotels, the popularity fusion classification might be: top 40% high, middle 30% medium, and bottom 30% low.

[0085] This embodiment introduces dynamic weight configuration associated with regional identifiers, enabling the calculation of heat fusion values ​​to adapt to the market channel characteristics and data distribution differences in different regions. This solves the evaluation bias problem caused by static weights and improves the regional adaptability and accuracy of the heat determination results. At the same time, the use of a preset weight configuration table to achieve dynamic weight reading ensures regional adaptability while reducing online calculation complexity and system resource consumption, thus improving the processing efficiency of the heat determination process.

[0086] Furthermore, in this embodiment, the weight values ​​in the weight configuration table are dynamically adjusted based on the analysis results of the influence of each data dimension on the popularity determination result in historical data. This embodiment allows the dynamic weights to not only adapt to regional characteristics but also to dynamically iterate with market changes and user behavior changes, further improving the accuracy and timeliness of popularity assessment. It is applicable to the maintenance of weight configuration tables in all regions, especially in rapidly changing overseas tourism markets (such as holidays and peak tourist seasons), enabling rapid adjustment of weights and ensuring that the popularity assessment results always align with market realities.

[0087] Specifically, historical data can be collected from various regions. This historical data includes raw data for each dimension of the evaluated object, evaluation parameters, popularity determination results, and market conversion data. For example, historical data for each region and each type of evaluation object can be collected using a streaming computing framework. The collection period can be set weekly or monthly (e.g., weekly during peak tourist seasons and monthly during off-seasons). The collection scope includes: raw data for each data dimension (output, traffic, reviews, exposure) and standardized evaluation parameters; historical popularity determination results (initial popularity level, target popularity level); market conversion data (order volume, transaction amount, traffic conversion rate, rating change rate, etc.); and market environment data (e.g., regional tourist numbers, consumption index, platform channel traffic changes). The collected historical data is cleaned, deduplicated, and correlated to establish a correlated dataset of regions, data dimensions, popularity results, and conversion data, ensuring the accuracy of the analysis results.

[0088] Then, historical data is analyzed to determine the degree of influence of each data dimension on the popularity determination result. The degree of influence refers to the correlation between changes in the evaluation parameters of a certain data dimension and changes in the popularity determination result (and market conversion data). The higher the correlation, the greater the influence of that dimension on the popularity determination, and the higher its weight should be assigned. Analysis methods can employ machine learning or statistical analysis methods such as correlation analysis, regression analysis, and feature importance assessment, including Pearson correlation coefficient and random forest feature importance scoring. Based on the results of the influence analysis, the weight values ​​of each data dimension corresponding to each region identifier are adjusted. The adjustment of weight values ​​should satisfy the principles of total conservation, unchanged primary and secondary elements, and dynamic adaptation. For example, the sum of the weight values ​​of the four major data dimensions is always 100%; the core weight status of output and reviews remains unchanged, and their combined weight is not less than 60% (a threshold can be set according to business needs); the weight values ​​of each dimension are fine-tuned based on changes in the influence score to avoid large fluctuations that could lead to unstable popularity results. Finally, the adjusted weight values ​​are updated to a preset weight configuration table to achieve dynamic iterative updates of the weight configuration table. Specifically, for the adjusted weight values, the weight configuration table is incrementally updated using the region identifier as an index, modifying only the weight values ​​of the regions and dimensions that have changed, while leaving the unchanged ones unchanged. If multiple regions exhibit the same market trend (such as increased local traffic in multiple countries in the Asia-Pacific region), batch updates can be performed for each region to improve update efficiency. After the update is completed, the latest weight configuration table is synchronized to the heat calculation system and downstream business systems in real time through message queues, API pushes, etc., ensuring that a unified weight configuration is used throughout the entire chain. Furthermore, historical versions of the weight configuration table can be retained, recording the time, reason, and analysis results of each adjustment, supporting weight rollback, and improving the system's fault tolerance.

[0089] In one possible implementation, the method further includes: determining the ranking of the evaluated object in each data dimension within its region based on the values ​​of the evaluation parameters in each data dimension; and correcting the evaluation parameters based on the ranking to obtain the corrected evaluation parameters.

[0090] The ranking of data dimensions within a region refers to the relative position obtained by ranking the evaluation parameters of a certain data dimension of the evaluated object in descending order with the evaluation parameters of all other evaluated objects in the same category within the same region. This relative position represents the regional market competitiveness of the evaluated object in that dimension. "Same category" refers to segmentation based on the type and positioning of the evaluated object (e.g., high-end hotels, budget guesthouses, scenic area shops) to avoid invalid comparisons between different categories and ensure the rationality of the ranking. Ranking correction involves weighting and integrating the original evaluation parameters of the evaluated object with the ranking percentage of that dimension to obtain corrected evaluation parameters. This corrected parameter retains the quantitative characteristics of the original data while incorporating the characteristics of relative regional competitiveness.

[0091] For example, the evaluation objects are grouped according to regional and category identifiers to ensure that objects within the same group are from the same region and category. For example, the first group is "Region A: Mid-to-high-end hotels," and the second group is "Region B: Budget homestays." For each group, the evaluation parameters for each data dimension (output, traffic, reviews, exposure) are sorted in descending order, with higher values ​​ranking higher. The ranking results are quantified by calculating the ranking percentage of the evaluation object in each dimension, using the formula: Ranking Percentage = (Number of objects after this object / Total number of objects in the group) × 100%. A higher ranking percentage indicates stronger regional competitiveness. The ranking and ranking percentage of the evaluation object in each data dimension are recorded as the core basis for parameter adjustment. If the total number of objects in a group is too small (e.g., <10), the ranking adjustment step is skipped, and the original evaluation parameters are used directly to avoid ranking distortion caused by insufficient sample size. Then, the original evaluation parameters of each data dimension can be corrected according to the preset ranking correction formula. For example, the preset ranking weight coefficient (denoted as α) is set, with a value range of 0.1-0.3 (which can be adjusted according to business needs). This represents the contribution of the ranking percentage in parameter correction. The weight coefficient of the original evaluation parameter is (1-α), ensuring that the original parameter is dominant. The ranking percentage is normalized and mapped to the same score range as the original evaluation parameter (e.g., if the original parameter is [0,10], then the ranking percentage (0-100%) is mapped to [0,10]), thus obtaining the ranking correction factor. The corrected evaluation parameter is calculated using the following formula: Corrected parameter = Original evaluation parameter × (1-α) + Ranking correction factor × α. The corrected parameter is subject to a threshold limit to ensure that its value is within the original score range (e.g., [0,10]), avoiding data distortion caused by exceeding the range.

[0092] This embodiment introduces a regional ranking-based evaluation parameter correction mechanism, eliminating the impact of differences in data distribution across different dimensions. This ensures that evaluation parameters for each dimension are comparable on the same relative scale, improving the fairness and rationality of subsequent weighted fusion. Simultaneously, ranking correction smooths out abnormal fluctuations in the original data, enhances the robustness of evaluation parameters, and more accurately reflects the actual competitive position of the evaluated object within the region, avoiding misjudgments of popularity due to a mismatch between absolute numerical values ​​and relative competitive position.

[0093] In one possible implementation, the process of determining the initial popularity value of an evaluation object based on the statistical data of review data for each reference evaluation object belonging to the first type includes:

[0094] Obtain the point ratings of each reference assessment object belonging to the first type; based on the point ratings of each reference assessment object, determine the mode, high popularity threshold, and low popularity threshold, where the mode is the most frequently occurring value among the point ratings of each reference assessment object, the high popularity threshold is the lower limit of the point ratings of reference assessment objects belonging to the first type and having a high popularity level, and the low popularity threshold is the upper limit of the point ratings of reference assessment objects belonging to the first type and having a low popularity level; obtain the point ratings of the assessment objects; based on the comparison results of the assessment object's point ratings with the mode, high popularity threshold, and low popularity threshold, determine the initial popularity level of the assessment object.

[0095] In this embodiment, for an evaluation object classified as Type II (with incomplete historical data), it is first necessary to obtain review data from all Type I evaluation objects (i.e., reference evaluation objects with complete data) belonging to the same region (such as the same country or the same city). Taking a hotel as an example, suppose a newly opened hotel is located in Region H and is classified as Type II. Then, the review scores of all Type I hotels (i.e., established hotels with complete historical data) in that region are obtained. These review scores can be the comprehensive scores of each hotel, such as the weighted and merged scores from multiple data platforms, and the value range is usually 0-10 or 0-5.

[0096] The mode is the most frequently occurring value in a dataset, representing the most typical rating level. For continuous data, the mode can be determined by grouping frequency statistics. The high-popularity threshold is the lower limit of the rating value for a reference assessment object belonging to the first category and having the highest popularity level. That is, among hotels with complete data, the minimum rating value a hotel classified as high-popularity must achieve. The low-popularity threshold is the upper limit of the rating value for a reference assessment object belonging to the first category and having the lowest popularity level. That is, among hotels with complete data, the maximum rating value a hotel classified as low-popularity must not exceed.

[0097] For evaluation objects classified as the second type (such as newly opened hotels), their existing review scores are obtained. This review data may be limited in quantity, but it can serve as the basis for the initial evaluation. The obtained review scores of the evaluation objects are compared with the three benchmark thresholds determined above, and their initial popularity level is determined according to preset judgment rules. Further, in this embodiment, the judgment rules can also be implemented in the following ways: Multi-level threshold division: The popularity level is subdivided into more levels (such as extremely high, high, medium-high, medium-low, low, etc.), and multiple threshold intervals are set accordingly. Weighted distance judgment: The weighted distance between the review score and the mode, high popularity threshold, and low popularity threshold is calculated, and the level with the closest distance is selected as the initial popularity level. Fuzzy matching: When the review score is close to the threshold boundary, a comprehensive judgment is made by combining other available data (such as exposure, traffic, etc.).

[0098] This embodiment uses review statistics of the first type of reference evaluation object to estimate popularity, solving the problem that the second type of evaluation object (cold start scenario) cannot be directly evaluated due to the lack of historical data. By utilizing the review distribution characteristics of hotels with complete data in the same area, a benchmark threshold system is constructed, enabling new hotels or hotels with missing data to obtain a reasonable initial popularity positioning, avoiding evaluation blind spots caused by insufficient data. Simultaneously, by employing multi-dimensional benchmarks such as the mode, high popularity threshold, and low popularity threshold, it takes into account typical levels, upper limits, and lower limits, making the estimation results more scientific and reasonable.

[0099] Furthermore, in this embodiment of the application, the process of determining the initial popularity level of the evaluation object based on the comparison results of the evaluation object's point score with the mode, high popularity threshold, and low popularity threshold includes:

[0100] If the rating of the evaluated object is equal to the mode, the initial popularity level of the evaluated object is determined to be the popularity level corresponding to the mode; if the rating of the evaluated object is not less than the high popularity threshold, the initial popularity level of the evaluated object is determined to be the high popularity level; if the rating of the evaluated object is not greater than the low popularity threshold, the initial popularity level of the evaluated object is determined to be the low popularity level; if the rating of the evaluated object is between the low popularity threshold and the high popularity threshold and is not equal to the mode, the initial popularity level of the evaluated object is determined to be the intermediate popularity level.

[0101] Specifically, a cold start refers to the problem of being unable to conduct evaluations using conventional data analysis methods for newly launched or historically scarce entities due to a lack of sufficient historical transaction records, user behavior data, etc. Let's take a hotel as an example.

[0102] When a cold-start hotel's review score is exactly equal to the mode of the review scores for Category 1 hotels in the region, it indicates that the hotel's review level is at the most typical level in that region. At this point, it's necessary to analyze beforehand which popularity range the mode falls within for Category 1 hotels. For example, statistical analysis might reveal that among Category 1 hotels in a certain region with a review score of 8.0 (the mode), 80% are classified as medium popularity, 15% as high popularity, and 5% as low popularity. Therefore, the popularity level corresponding to the mode can be determined as medium popularity. Thus, a cold-start hotel with a review score of 8.0 has its initial popularity level determined as medium popularity. When a cold-start hotel's review score is greater than or equal to the high popularity threshold (i.e., the minimum review score for high popularity hotels in Category 1), it indicates that the hotel's review level has reached or even exceeded the threshold for high popularity hotels in a mature market. For example, if the high popularity threshold for a certain region is 8.5, and a newly opened hotel has a review score of 8.7, meeting the condition of not less than 8.5, its initial popularity level is directly determined as high popularity. This rule ensures that cold-start hotels with a small amount of data but excellent reputation can obtain popularity positioning that matches their quality.

[0103] When a cold-start hotel's review score is less than or equal to the low-popularity threshold (i.e., the highest review score for low-popularity hotels in the first category), it indicates that the hotel's review level falls within the range of low-popularity hotels in a mature market. For example, if the low-popularity threshold for a certain region is 5.5, and a newly opened hotel has a review score of 4.8, it meets the condition of not exceeding 5.5, therefore its initial popularity level is directly determined as low-popularity. This rule can effectively identify cold-start hotels with poor reputation performance and prevent them from obtaining excessively high popularity positioning that does not match their actual quality. When a cold-start hotel's review score is between the low-popularity threshold and the high-popularity threshold, and is not equal to the mode, it indicates that the hotel's review level is in the middle of a mature market, neither reaching the high-popularity threshold nor falling into the low-popularity range, and not belonging to the most typical mode level. In this case, its initial popularity level is determined as intermediate popularity level. For example, a region might have a low popularity threshold of 5.5 points, a high popularity threshold of 8.5 points, and a mode of 8.0 points. A newly opened hotel might have a rating of 7.0 points, satisfying the condition that 5.5 < 7.0 < 8.5 and 7.0 ≠ 8.0. Therefore, its initial popularity level is considered to be intermediate. Through these judgment rules, this embodiment achieves accurate and efficient estimation of the initial popularity level of the second type of assessment object, providing a reliable input basis for subsequent correction processing.

[0104] In one possible implementation, the process of correcting the initial popularity level to obtain the target popularity level of the evaluation object includes: acquiring the attribute data of the evaluation object, wherein the attribute data includes at least one of region identifier, location type, and service capacity data; matching the attribute data with condition items in a preset constraint condition library, wherein the constraint condition library contains at least one constraint condition, and each constraint condition includes a condition item and a corresponding correction operation; if the match is successful, adjusting the initial popularity level according to the correction operation corresponding to the matched target constraint condition to obtain the target popularity level.

[0105] Among them, attribute data refers to the basic characteristic data of the assessed object in addition to multi-source heterogeneous operational data. It serves as the basis for revising business rules and reflects the core business attributes of the assessed object, such as its regional affiliation, business type, and service capabilities. It is strongly correlated with the actual business value of the popularity level. Venue type refers to the business category / type of the assessed object, such as hotels, guesthouses, boutique inns, and vacation apartments. Different types correspond to different operational rules. Service capacity data refers to the core service carrying capacity indicators of the assessed object, such as the number of rooms for hotels / guesthouses, the number of visitors for scenic spots, and the operating area for shops. It is an indicator for measuring the operational scale of the assessed object.

[0106] In this embodiment, before performing correction processing, it is first necessary to obtain the attribute data of the evaluation object. This attribute data reflects the inherent characteristics of the evaluation object itself and, unlike multi-source heterogeneous data, does not change frequently over time. Taking a hotel as an example, the regional identifier can be the country or city where the hotel is located, such as "Country A, Region S"; the location type can be "hotel", "guesthouse", "apartment", etc.; the service capacity data can be "number of rooms", such as 45 rooms, 200 rooms, etc. This attribute data can usually be directly read from the hotel's basic information database. A constraint condition library is pre-built to store correction rules set based on business experience or operational strategies. Each constraint condition consists of two parts: a condition item and a correction operation. The condition item limits the attribute data, and the correction operation is the specific adjustment method for the initial popularity level. The constraint condition library can be stored in the database in the form of a rule table, supporting dynamic addition, deletion, modification, and query. For example, if the evaluation object is a hotel, the condition item is service capacity data ≥ 500 rooms and the initial popularity level is medium popularity, and the correction operation is to adjust the initial popularity level to high popularity.

[0107] The acquired attribute data is matched one by one with the conditions in the constraint library. Taking a newly opened guesthouse in area A as an example, its attribute data is: area identifier "A00", location type "guesthouse", and service capacity of 45 rooms. The initial popularity level is medium popularity. This attribute data is then matched with the constraint library:

[0108] The first condition is met: the area identifier is "A00", the venue type is "homestay", and the service capacity data is ≤50 rooms. The match is successful. Based on the matched correction operation "adjust the initial popularity level to low popularity", the hotel's initial popularity level is adjusted from medium popularity to low popularity, thus obtaining the target popularity level.

[0109] If the attribute data does not match any condition in the constraint library, the initial popularity level remains unchanged and is directly used as the target popularity level. For example, if a hotel area is identified as "B01", the venue type is "hotel", the service capacity is 300 rooms, and the initial popularity level is high, and no condition is matched, its target popularity level remains high. Through the above correction process, this embodiment achieves the fusion of the data model output results and preset constraints, so that the final popularity level is both based on multi-source data calculation and meets the special requirements of actual business scenarios, improving the practicality and acceptability of the popularity determination results.

[0110] Furthermore, in this embodiment of the application, the constraints in the constraint library include at least one of the following:

[0111] If the area is identified as the first preset area and the venue type is a preset type, and the service capacity data is lower than the first capacity threshold, the initial popularity level will be adjusted to a low popularity level; if the service capacity data is higher than the second capacity threshold and the initial popularity level is a medium popularity level, the initial popularity level will be adjusted to a high popularity level; if the service capacity data is lower than the third capacity threshold and the initial popularity level is a high popularity level, the initial popularity level will be adjusted to a medium popularity level.

[0112] For example, when the evaluation object's region is identified as a first preset region (e.g., South Korea), its location type is a preset type (e.g., guesthouse), and its service capacity data is lower than a first capacity threshold (e.g., 50 rooms), the initial popularity level is adjusted to a low popularity level. Taking a guesthouse in Seoul, South Korea as an example, with 45 rooms, its initial popularity level is medium popularity. Because it meets the condition of "South Korea + Guesthouse + Number of Rooms ≤ 50," a correction operation is triggered, adjusting its popularity level from medium popularity to low popularity. This constraint is based on the following considerations: In the South Korean guesthouse market, properties with too few rooms typically have limited accommodation capacity. Even with a good reputation, it is not advisable to assign too high a popularity level, lest insufficient accommodation capacity affect service capacity.

[0113] When the service capacity data of an assessed entity exceeds the second capacity threshold (e.g., 500 rooms) and its initial popularity level is intermediate, the initial popularity level is adjusted to high. For example, a large resort hotel with 800 rooms initially has an intermediate popularity level. Because it meets the condition of "number of rooms ≥ 500 and initial popularity is intermediate," a correction operation is triggered, adjusting its popularity level from intermediate to high. This constraint is based on the following considerations: hotels with a large number of rooms typically have higher market stock and operational complexity, and their market influence often exceeds that of smaller hotels with similar reputations. Appropriately increasing the popularity level allows resource allocation to match the scale of the entity.

[0114] When the service capacity data of an assessed property is below the third capacity threshold (e.g., 30 rooms) and its initial popularity level is high, the initial popularity level is adjusted to a medium popularity level. For example, a boutique guesthouse has 25 rooms, but due to its excellent reputation, its initial popularity level is high. Since it meets the condition of "number of rooms ≤ 30 and initial popularity is high," a correction operation is triggered, adjusting its popularity level from high to medium. This constraint is based on the following considerations: even if a property has a good reputation, its actual reception capacity and market influence are limited if it has too few rooms. Assigning it too high a popularity level may lead to excessive resource allocation with limited actual benefits; appropriate downgrading can make resource allocation more reasonable.

[0115] This embodiment corrects the initial popularity level through preset constraints, ensuring that the popularity determination result is compatible with the inherent attribute characteristics of the evaluation object, thus resolving the disconnect between pure data model output and actual conditions. By employing a rule-based approach where condition items correspond to correction operations, business experience can be embedded into the data processing flow in a structured manner, improving the interpretability and acceptability of the popularity determination result. Furthermore, the constraint library supports dynamic expansion, allowing for flexible configuration of rules based on the specific needs of different regions and types of evaluation objects, enhancing the adaptability and practicality of the popularity determination method.

[0116] In one possible implementation, the method further includes: outputting the target popularity level to a downstream system so that the downstream system performs corresponding data processing operations on the evaluation object according to the target popularity level.

[0117] When outputting the target popularity level to downstream systems, the output can be performed according to a preset method. For example, digital transmission can meet the requirements of real-time performance, high availability, and compatibility, supporting millisecond-level synchronization. Downstream systems refer to the core business systems that receive the target popularity level and execute specific business decisions. They are the application carriers of the popularity assessment results. For example, when the assessment object is a hotel, downstream systems may include systems for resource allocation, traffic scheduling, marketing promotion, and room status management in overseas hotel operations. In addition to the target popularity level, the output content also includes the unique identifier of the assessment object, regional identifier, update time, and other relevant information to ensure that downstream systems can accurately associate with the corresponding assessment object.

[0118] First, determine the standardized output format: output data should use lightweight formats such as JSON / Protobuf, including core fields such as the evaluation object ID, region identifier, target popularity level (high / medium / low), popularity update time, and data version number, ensuring downstream systems can parse it quickly. Choose a high real-time output method: select the corresponding output method based on the downstream system's technical architecture. The core methods include two types: real-time interface push, where the popularity evaluation system provides a standardized RESTful / GRPC interface, and downstream systems actively pull or push the target popularity level, with a response time ≤200ms; and message queue synchronization, where the target popularity level is sent to message queues such as Kafka / RabbitMQ, and downstream systems subscribe to queue messages to achieve asynchronous real-time synchronization, avoiding system coupling. Perform batch output to multiple downstream systems, simultaneously outputting the target popularity level to all relevant downstream business systems to ensure consistent popularity data across systems and avoid data silos. Add an output fault tolerance mechanism, automatically retrying (e.g., 3 retries) for output failures, storing failed retry records offline for later re-push, ensuring data integrity. Record output logs, including output time, downstream systems, output results, version number, and other information for each evaluation object, to support subsequent traceability and problem investigation.

[0119] After receiving the target popularity level, the downstream system performs differentiated data processing operations on the evaluated object according to a preset strategy. The following explanation uses the following downstream system type as an example: When the downstream system is a search ranking system, it can dynamically adjust the hotel's ranking weight in search results based on its popularity level: High-popularity hotels: ranking weight increased by 20%-30%, prioritized display in search results; Medium-popularity hotels: maintain default ranking weight, displayed normally; Low-popularity hotels: ranking weight decreased by 10%-20%, displayed appropriately later. When the downstream system is a recommendation system, it can determine the recommendation frequency and exposure position based on the hotel's popularity level: High-popularity hotels receive increased exposure in homepage recommendations and personalized recommendation streams; Medium-popularity hotels are displayed normally in related recommendations and similar hotels; Low-popularity hotels have reduced recommendation frequency or are only displayed under specific filtering conditions. When the downstream system is an operations management system, it can allocate human operational resources according to the hotel's popularity level: high-popularity hotels can be equipped with dedicated operations personnel to prioritize handling needs such as room status synchronization and price adjustments; medium-popularity hotels can receive routine operational support and periodically monitor data changes; low-popularity hotels can reduce human intervention and perform basic maintenance through automated processes.

[0120] This embodiment achieves rapid linkage between the heat determination result and the business processing link by outputting the target heat level to the downstream system in real time in the form of streaming data, compressing the response delay to the millisecond level, and enabling the downstream system to perform differentiated data processing operations in a timely manner according to the heat level.

[0121] The following example illustrates the method for determining the popularity of the evaluation object in this application, using a practical application scenario as an example.

[0122] Referring to Figure 2, it illustrates a flowchart of a heat map processing method for an evaluation object provided in an embodiment of this application. This embodiment uses the hotel or guesthouse market in Seoul, South Korea as an application scenario, combined with... Figure 2 The process is described in detail, outlining the execution of the popularity stratification. First, the basic indicator scoring step is executed. Corresponding to the multi-source data fusion and preprocessing stage of this application, the heterogeneous data from multiple sources across four dimensions—output, traffic, reviews, and exposure—are standardized to obtain basic indicator scores for each dimension. Next, the "Is it a core hotel?" judgment node is entered, corresponding to the evaluation object type classification stage. Based on the completeness of historical data, it is determined whether the hotel is a core hotel (Type 1): if the number of historical data records for all four dimensions reaches the target number range for the Seoul market, it is determined to be a core hotel; if at least one dimension's data does not meet the standard, it is determined to be a non-core hotel.

[0123] For core hotels, the core merchant stratification process proceeds as follows: First, the steps of calculating the integrated rating score, weighted rating score, and unweighted rating score are executed, corresponding to the core hotel weighted integrated assessment stage. A weighted rating score is calculated based on dynamic weights, and an unweighted rating score and integrated rating score are generated simultaneously. The highest of the three is taken as the merchant's score. Next, the core merchant stratification step is executed, corresponding to the core hotel initial popularity level determination stage. Based on the comparison between the merchant's score and a preset level threshold, the initial stratification of core merchants is determined. For non-core hotels, the non-core merchant stratification process proceeds as follows: First, the steps of calculating the national review mode, highest review score, and lowest review score are executed, corresponding to the non-core hotel market benchmark construction stage. Based on review data of core hotels in the Seoul area, the review mode, highest popularity threshold (high), and lowest popularity threshold (low) are calculated. Next, the non-core merchant stratification step is executed, corresponding to the non-core hotel initial popularity level determination stage. The review scores of non-core hotels are compared with the market benchmark to determine the initial stratification. After completing the stratification of core or non-core merchants, the process proceeds to the "Does the number of rooms meet the standard?" judgment node. This corresponds to the business rule modification stage of this application, where the service capacity (number of rooms) of the evaluated object is used to determine whether to trigger a modification operation. If the number of rooms meets the preset room quantity rule conditions (e.g., homestay ≤ 50 rooms, hotel > 500 rooms), the "Adjust merchant stratification according to different scenarios" step is executed to adjust the initial stratification; if the rule conditions are not met, the "Maintain original stratification" step is executed to maintain the initial stratification result, ultimately completing the popularity stratification process.

[0124] pass Figure 2 The process shown enables differentiated popularity stratification for core and non-core merchants, ensuring the accuracy of core hotel evaluations with complete data, providing an implementation solution for hotels starting up from scratch, and making the stratification results more aligned with actual business scenarios through room volume rule adjustments, effectively reducing the risk of resource mismatch and enhancing the guiding value for operational decisions.

[0125] See Figure 3This document illustrates a flowchart of a data fusion and evaluation process for assessing the popularity of an evaluation object, as provided in an embodiment of this application. This embodiment uses the evaluation of the popularity of overseas hotels or guesthouses as an application scenario. First, a multi-source data collection step is performed, corresponding to the multi-source heterogeneous data acquisition stage of this application. This involves collecting multi-source data such as internal reviews, internal output, and internal and external traffic from channels including internal reviews, global search engines, and international or local tourism platforms. Simultaneously, external data is supplemented from dimensions such as global and vertical media exposure, relevant platform inventory (e.g., the traffic of the platform currently being evaluated), market share performance in specific regions, and comprehensive reputation scores, forming a comprehensive data foundation. It should be noted that the data collected in this embodiment are all publicly available data from public platforms. Next, the basic indicator calculation step is performed, corresponding to the multi-source data fusion and preprocessing stage of this application. The collected multi-source data is standardized to generate basic indicator scores and popularity scores, eliminating the differences in the dimensions of different data dimensions and providing a unified quantitative basis for subsequent stratification.

[0126] After completing the calculation of basic indicators, the process proceeds to the "Distinguish between core merchants?" decision point. This corresponds to the assessment object type classification stage, where the completeness of historical data is used to determine whether a merchant is a core merchant: if the hotel or guesthouse meets the target number of historical data records in all data dimensions, it is determined to be a core merchant; if at least one dimension's data does not meet the standard, it is determined to be a non-core merchant. For core merchants, the core merchant popularity strategy steps are executed, corresponding to the core hotel refined weighted assessment stage. Based on dynamic weights, the basic indicator scores are weighted and integrated to generate the core merchant popularity strategy results. For non-core merchants, the non-core merchant popularity strategy steps are executed, corresponding to the cold start hotel market benchmarking stage. Based on the statistical volume of review data of core merchants in the same region, the non-core merchant popularity strategy results are generated. Finally, the comprehensive measurement final result step is executed. Corresponding to the heat level determination and real-time update stage of this application, the heat strategy results of core or non-core merchants are comprehensively measured to generate the final heat evaluation result. The result is synchronized and updated in real time through a streaming update mechanism to ensure that the heat results can quickly reach the downstream business system and provide support for resource allocation, traffic scheduling and other decisions.

[0127] See Figure 4 This document illustrates an application scenario and output result architecture diagram for hotel popularity metrics provided in an embodiment of this application. This embodiment uses overseas hotel operation decision-making as an application scenario, combined with... Figure 4 The architecture fully explains the application process of hotel popularity assessment results. This architecture outputs two core results: first, popularity stratification, used to classify individual hotels into high, medium, and low popularity levels; second, popularity score, used to generate a quantitative comprehensive popularity rating, providing a refined decision-making basis for downstream business.

[0128] In the process of processing popularity stratification, the output results can be directly applied to multiple core business scenarios: for popular hotel recommendations, prioritizing the display of high-popularity hotels in the platform's homepage recommendation position; for the hotel best-of-life database (EBK) admission, including high-popularity hotels in the preferred resource pool, allocating high-quality traffic and human operation resources; for hotel dimension table updates, synchronizing the popularity stratification results to the hotel basic information database, serving as the core basis for hotel tags and operational priorities, and can also be extended to more business scenarios (such as marketing campaign admission, room status management priority, etc.).

[0129] In the process of processing popularity scores, the output results can serve as core inputs for multi-dimensional decision-making: for calculating city popularity scores, aggregating the popularity scores of individual hotels into a city-level popularity index to guide regional resource allocation and market layout; for adjusting price comparison weights, dynamically adjusting price comparison strategies based on hotel popularity scores, with high-popularity hotels receiving higher price comparison priority; and for setting service assessment weights, incorporating popularity scores into the hotel service assessment system as a core reference for assessment weights, and extending to more refined decision-making scenarios (such as marketing subsidy amounts, room availability synchronization frequency, etc.).

[0130] pass Figure 4 The architecture shown enables multi-scenario and multi-dimensional applications of hotel popularity assessment results. It achieves refined data processing and accurate resource allocation for individual hotels through popularity stratification, and supports city-level and strategy-level business decisions through popularity scores. This transforms the popularity assessment results from a single evaluation indicator into a core basis for business decision-making covering the entire chain, effectively improving the level of refinement and resource utilization efficiency of overseas hotel operations.

[0131] This application embodiment also provides a heat determination device for an evaluation object, the device comprising:

[0132] The data acquisition unit is used to acquire multi-source heterogeneous data of the evaluation object, wherein the multi-source heterogeneous data includes at least production data, traffic data, review data and exposure data;

[0133] The first determining unit is used to determine the evaluation parameters of the evaluation object in each data dimension based on the multi-source heterogeneous data.

[0134] The second determining unit is used to determine the target type of the evaluation object based on the completeness of the historical data of the evaluation object. The target type includes a first type and a second type. The first type indicates that the number of historical data records of the evaluation object in each data dimension has reached the target number range. The second type indicates that the number of historical data records of the evaluation object in at least one data dimension has not reached the target number range.

[0135] The first processing unit is configured to, in response to the target type of the evaluation object being a first type, perform weighted fusion processing on the evaluation parameters based on dynamic weights to obtain the initial popularity level of the evaluation object. The dynamic weights characterize the contribution of evaluation parameters of different data dimensions in the popularity determination process, and the value of the dynamic weights is related to the region to which the evaluation object belongs.

[0136] The third determining unit is used to determine the initial popularity level of the evaluation object based on the statistical data of the review data of each reference evaluation object belonging to the first type, in response to the target type of the evaluation object being the second type.

[0137] The second processing unit is used to correct the initial heat level to obtain the target heat level of the evaluation object.

[0138] It should be noted that the various units and their specific implementations in this embodiment can be referred to the corresponding content above, and will not be described in detail here.

[0139] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the methods for determining the heat of an evaluation object provided in this application.

[0140] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the methods for determining the heat of an evaluation object provided in this application.

[0141] This application also provides an electronic device, including at least one processor and a memory connected to the processor, wherein: the memory is used to store a computer program; the processor is used to execute the computer program so that the electronic device can implement any of the methods for determining the heat of an evaluation object provided in this application.

[0142] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0143] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0144] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0145] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining the popularity of an evaluation object, characterized in that, include: Obtain multi-source heterogeneous data of the evaluation object, wherein the multi-source heterogeneous data includes at least production data, traffic data, review data, and exposure data; Based on the multi-source heterogeneous data, the evaluation parameters of the evaluation object in each data dimension are determined; Based on the completeness of the historical data of the evaluation object, the target type of the evaluation object is determined. The target type includes a first type and a second type. The first type indicates that the number of historical data records of the evaluation object in each data dimension has reached the target number range. The second type indicates that the number of historical data records of the evaluation object in at least one data dimension has not reached the target number range. In response to the target type of the evaluation object being the first type, the evaluation parameters are weighted and fused based on dynamic weights to obtain the initial popularity level of the evaluation object. The dynamic weights represent the contribution of evaluation parameters of different data dimensions in the popularity determination process, and the value of the dynamic weights is related to the region to which the evaluation object belongs. In response to the target type of the evaluation object being the second type, the initial popularity level of the evaluation object is determined based on the statistical data of the review data of each reference evaluation object belonging to the first type; The initial heat level is corrected to obtain the target heat level of the evaluation object.

2. The method according to claim 1, characterized in that, The step of weighting and fusing the evaluation parameters based on dynamic weights to obtain the initial popularity level of the evaluation object includes: Obtain the region identifier of the area to which the evaluation object belongs; Based on the region identifier, the weight values ​​of each data dimension corresponding to the region identifier are read from a preset weight configuration table, wherein the weight values ​​of each data dimension corresponding to different region identifiers are at least partially different. Based on the weight values ​​of each data dimension, the evaluation parameters are weighted and summed to obtain the heat fusion value; The initial popularity level of the evaluation object is determined based on the comparison result between the popularity fusion value and the preset level classification threshold.

3. The method according to claim 2, characterized in that, The weight values ​​in the weight configuration table are dynamically adjusted based on the analysis results of the influence of each data dimension on the determination of popularity in historical data.

4. The method according to claim 1, characterized in that, The method further includes: Based on the values ​​of the evaluation parameters in each data dimension, determine the ranking of the evaluation object in each data dimension within its region; The evaluation parameters are adjusted based on the ranking to obtain the adjusted evaluation parameters.

5. The method according to claim 1, characterized in that, The determination of the initial popularity level of the evaluation object based on the statistical data of each reference evaluation object belonging to the first type includes: Obtain the point ratings for each reference assessment object belonging to the first type; Based on the point ratings of each reference evaluation object, the mode, high popularity threshold, and low popularity threshold are determined. The mode is the most frequently occurring value among the point ratings of each reference evaluation object. The high popularity threshold is the lower limit of the point rating of reference evaluation objects belonging to the first type and having a high popularity level. The low popularity threshold is the upper limit of the point rating of reference evaluation objects belonging to the first type and having a low popularity level. Obtain the point rating value of the evaluation object; The initial popularity level of the evaluated object is determined based on the comparison results between the point score of the evaluated object and the mode, high popularity threshold and low popularity threshold.

6. The method according to claim 1, characterized in that, The step of determining the initial popularity level of the evaluation object based on the comparison results between the evaluation object's point score and the mode, high popularity threshold, and low popularity threshold includes: If the rating of the evaluated object is equal to the mode, the initial popularity level of the evaluated object is determined to be the popularity level corresponding to the mode; If the score of the evaluated object is not less than the high popularity threshold, the initial popularity level of the evaluated object is determined to be a high popularity level; If the point score of the evaluated object is not greater than the low popularity threshold, the initial popularity level of the evaluated object is determined to be low popularity level; If the rating of the evaluated object is between the low popularity threshold and the high popularity threshold and is not equal to the mode, the initial popularity level of the evaluated object is determined to be the intermediate popularity level.

7. The method according to claim 1, characterized in that, The step of correcting the initial popularity level to obtain the target popularity level of the evaluation object includes: Obtain attribute data of the assessment object, wherein the attribute data includes at least one of the following: area identifier, location type, and service capacity data; The attribute data is matched with the condition items in the preset constraint condition library. The constraint condition library contains at least one constraint condition, and each constraint condition includes a condition item and a corresponding correction operation. If a match is successful, the initial heat level is adjusted according to the correction operation corresponding to the matched target constraint to obtain the target heat level.

8. The method according to claim 7, characterized in that, The constraints in the constraint library include at least one of the following: If the area is identified as a first preset area, the venue type is a preset type, and the service capacity data is lower than the first capacity threshold, the initial popularity level will be adjusted to a low popularity level. If the service capacity data is higher than the second capacity threshold and the initial popularity level is the intermediate popularity level, the initial popularity level will be adjusted to the high popularity level. If the service capacity data is lower than the third capacity threshold and the initial popularity level is high popularity level, then the initial popularity level will be adjusted to the intermediate popularity level.

9. The method according to claim 1, characterized in that, Also includes: The target popularity level is output to the downstream system so that the downstream system can perform corresponding data processing operations on the evaluation object according to the target popularity level.

10. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the heat determination method for the evaluation object as described in any one of claims 1-9.

Images