Cosmetic efficacy evaluation method and system based on efficacy semantic mapping

Abstract

The application discloses a cosmetic efficacy evaluation method and system based on efficacy semantic mapping, and relates to the technical field of data processing of human efficacy evaluation of cosmetics; heterogeneous measurement records containing original field names, instrument identifiers, units, time points and measurement values are received, candidate standard index domains are constructed based on instrument measurable index sets and scheme target index sets, and scores are given according to field texts, instrument compatibility, unit compatibility, scheme context and numerical characteristics; unique standard indexes or conflict and unidentified states are determined by using a receiving threshold and a candidate interval threshold; only the mapping confidence and the mapping version corresponding to the unique standard indexes are written into the records and addressed to determine rules, time point evidences, single-index cross-time point evidences and single-item efficacy cross-index evidences are formed, and finally, a traceable evidence chain associated with original records, rule versions and report contents is generated.

Description

Technical Field

[0001] This invention relates to the field of data processing technology for cosmetic efficacy evaluation on human body, and more specifically, to a method and system for evaluating cosmetic efficacy based on efficacy semantic mapping. Background Technology

[0002] Cosmetic efficacy evaluation projects typically include protocol configuration, subject management, multiple visit time points, various instrument measurements, statistical analysis, and report generation. Different testing instruments and their export software may use different field names, abbreviations, units, and channel expressions for the same evaluation parameter, and the same efficacy claim often requires multiple standard evaluation indicators to support it.

[0003] Existing technologies typically employ fixed fields, manual mapping, or generic string similarity matching to process instrument data. Fixed fields require simultaneous modifications to the table structure or program when new instruments and non-standard parameters are added; manual mapping involves a large amount of repetitive work in projects with multiple instruments and multiple time points; and string-based matching alone struggles to distinguish between short and similar indicators.

[0004] The field name matching process does not simultaneously utilize technical contexts such as instrument capabilities, units, and program objectives, which may lead to the selection of seemingly similar candidate indicators; furthermore, the uncertainty in field interpretation is usually not included in the statistical calculation branch, making it difficult to prevent unreliable records from affecting the final efficacy conclusions.

[0005] Existing solutions often maintain data entry fields, statistical rules, and reporting variables separately. Even if an original field is misinterpreted as another standard indicator, the system may still continue to select incorrect directions of change, units, and thresholds, ultimately leading to cross-stage semantic inconsistencies between statistical conclusions and report content. After rules or alias libraries are updated, if the versions used in historical records are not fixed, the same original data may yield different recalculation conclusions.

[0006] To address this, a cosmetic efficacy evaluation method and system based on efficacy semantic mapping is proposed. Summary of the Invention

[0007] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a method and system for evaluating the efficacy of cosmetics based on efficacy semantic mapping.

[0008] To achieve the above objectives, the present invention provides the following technical solution: receiving efficacy claims and heterogeneous measurement records; determining the set of target standard indicators for the scheme based on efficacy-standard indicator relationships and scheme-level coverage relationships; establishing candidate standard indicator domains by combining instrument-indicator relationships; performing joint scoring on the original field names in the candidate domains based on text, instrument, unit, scheme context, and numerical or channel characteristics, and classifying the records into uniquely determined, conflicting, and unidentified states using a receiving threshold and a candidate interval threshold.

[0009] Only the standard indicator identifier, mapping confidence level, and mapping version under a uniquely determined state are written into the measurement record. The standard indicator identifier is the control key for subsequent processing: its direct addressing direction, change threshold, stability limit, unit compatibility, and data quality conditions; the mapping confidence level and data quality jointly determine whether time-point evidence can be automatically upgraded to strong evidence; evidence from different time points is then successively summarized into single-indicator cross-time-point evidence and single-item efficacy cross-indicator evidence.

[0010] Specifically, before standardization, the heterogeneous measurement records save the subject identifier, protocol identifier, instrument identifier, original field name, unit identifier, time point, sample identifier, and one or more measurement values. When parallel measurements exist, the mean, standard deviation, coefficient of variation, acquisition channel, and quality control status are further saved. After standardization, standard indicator identifiers, mapping status, mapping confidence level, candidate list, mapping version, and judgment rule version are appended to the same record to ensure a one-to-one correspondence between the original record and the data interpretation results.

[0011] Specifically, system modules can be mirrored with method steps: the scheme semantic module determines the set of target standard indicators for the scheme; the record receiving module retains the original measurement records along with their instrument, unit, and time point context; the candidate domain construction module outputs constrained candidate standard indicator domains; the conflict resolution module outputs uniquely determined, conflicting, or unidentified states; the rule addressing module reads the corresponding indicator judgment rule using a uniquely determined standard indicator identifier; the evidence summarization module outputs evidence at each level; and the evidence chain output module projects the version-fixed results onto the report context. Modules pass outputs from the previous module, avoiding simply replacing method steps with names.

[0012] Specifically, the candidate standard indicator domain is preferentially selected from the intersection of the instrument's measurable indicator set and the scheme's target standard indicator set. When the intersection is empty, a fallback set with scoring penalties can be used to avoid unconstrained use of surface approximations from the global alias library for the current instrument or scheme. The original fields, instrument, unit, candidate list, mapping status, confidence level, and mapping version constitute a reproducible mapping result.

[0013] Specifically, candidate scores can be obtained by a weighted combination of text matching, instrument compatibility, unit compatibility, scheme context, and numerical range or channel compatibility. A dual-threshold conflict rejection mechanism is employed: a unique standard indicator is output only when the highest score reaches the acceptance threshold and the difference between it and the second-highest score reaches the interval threshold; otherwise, a conflict or unrecognized status is output, and the record is not allowed to automatically trigger efficacy rule addressing.

[0014] Specifically, directional indicators form time-point evidence based on significance, expected direction, and minimum magnitude of change; stable indicators form time-point evidence based on stability limits or equivalent intervals. When the mapping confidence level or data quality is below the threshold, the time point at most forms weak evidence, thus ensuring that the reliability of field standardization truly enters the calculation path of statistical conclusions.

[0015] Specifically, for single-indicator cross-time point evidence, the evidence values ​​corresponding to strong, weak, and invalid evidence are fused using preset time point weights; for single-efficacy cross-indicator evidence, the single-indicator cross-time point evidence of the associated standard indicator is fused using preset indicator weights and mapping confidence levels. When the sum of the effective weights participating in the fusion is zero, there is strong reverse evidence for the main evaluation indicator, or the main evaluation indicator does not meet the minimum evidence requirements, the system outputs no data, downgrade, or manual review status, instead of replacing it with the default score.

[0016] Specifically, each time an efficacy-standard indicator relationship, instrument-indicator relationship, indicator alias relationship, or indicator judgment rule is released, an independent rule version is generated, and a hash fingerprint is configured for the current version. Historical results continue to reference the mapping version and judgment rule version at the time of their formation by default; when recalculated according to the new rules, a new result version is generated that coexists with the old results. The report content only reads the evidence chain object of the solidified version and does not reinterpret the original fields during the template filling stage.

[0017] Specifically, the reporting module organizes the protocol, subjects, standardized measurement records, and statistical results into a unified reporting context. For common variable placeholders, the module writes corresponding values ​​or text based on the variable mapping table; for dynamic content blocks, the module generates statistical tables, time-point evidence, and efficacy conclusions based on the established chain of evidence. When no available master template is provided, the module can use its built-in report generator to output a structured document. Regardless of the master template used, the indicator names, units, efficacy attribution, and conclusions in the report sections all reference the same standard indicator identifier and its versioned evidence, avoiding manual mapping again during the reporting stage.

[0018] Specifically, there are prior- and subsequent data dependencies among the candidate domains, scoring items, dual thresholds, rule versions, and report projections of heterogeneous measurement records: if a candidate domain is not jointly constrained by the instrument and the protocol, erroneous candidates will be included in the scoring; if conflicting records are not rejected, erroneous standard indicators will address erroneous rules; if mapping confidence is not involved in evidence gating, low-confidence records may form strong evidence; if the report does not cite a fixed version, rule updates may cause historical conclusions to drift. Thus, these technical features work together to form reproducible efficacy evaluation results.

[0019] Specifically, time-point weights are used to fuse time-point evidence into single-indicator cross-time-point evidence, and indicator weights and mapping confidence scores are used to fuse related indicators into single-item efficacy cross-indicator evidence. The final evidence chain links the original measurement records, standard indicators, mapping versions, judgment rule versions, time-point statistics, evidence at all levels, and report outputs.

[0020] The technical effects and advantages of this invention are as follows: By narrowing the candidate domain together with the instrument set and scheme set, and automatically confirming it with joint scoring and dual threshold rejection restrictions, it can block erroneous index mapping when the original field is short, approximate, or incomplete in context, and prevent erroneous units, directions, and thresholds from entering subsequent rules.

[0021] By linking unique standard indicators, mapping confidence scores with rule addressing and evidence gating, the uncertainty of field identification can be transformed into a clear statistical processing branch, preventing low-confidence data from being automatically written as strong power conclusions.

[0022] By solidifying the mapping version and the judgment rule version, and having the reporting module consume only the solidified evidence chain objects, historical reports can follow the data interpretation path at the time of their issuance, thereby enabling the efficacy conclusions to be traced back from the report content to the original instrument records. Attached Figure Description

[0023] Figure 1 This is a flowchart of a cosmetic efficacy evaluation method based on efficacy semantic mapping according to the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] like Figure 1 As shown, a cosmetic efficacy evaluation method and system based on efficacy semantic mapping uses standard indicator identifiers as unique semantic keys between data structures. These standard indicator identifiers are not display names, but rather control objects passed between heterogeneous raw records, rule configurations, statistical evidence, and report content. The system can be deployed on servers, workstations, or computing devices that communicate with cosmetic human efficacy evaluation instruments.

[0026] The system maintains the following relationships: Efficiency-Standard Indicator Relationship (EMR), Instrument-Indicator Relationship (IMR), Indicator Alias ​​Relationship (MAR), Indicator Judgment Rule (MJR), and Protocol-Level Coverage Relationship (POR). EMR records the correspondence between efficacy claims and the default set of standard indicators; IMR records the standard indicators that a specified instrument can output; MAR records the Chinese abbreviation, full English name, English abbreviation, synonyms, and instrument field aliases for the standard indicators; MJR records the internal keys, expected direction of change, thresholds, stability limits, unit compatibility conditions, quality conditions, and rule versions for the standard indicators; and POR is used to limit the addition, deletion, or replacement of default efficacy-indicator relationships by a specific evaluation protocol.

[0027] In step S1, the efficacy claims of the proposed treatment are received, and the default standard indicator set is read from the EMR. If the current treatment has a Proof of Responsibility (POR), the POR is applied to the default standard indicator set to obtain the treatment's target standard indicator set. The target standard indicator set is used not only to select the evaluation indicators to be collected or statistically analyzed, but also as a contextual constraint during subsequent standardization of the original fields. For example, for a moisturizing treatment, this set may include skin moisture content and transepidermal water loss (TEWL); for a repair treatment, this set may also include a Values ​​and skin elasticity R2.

[0028] In step S2, heterogeneous measurement records are received. Each record includes at least the original field name x, instrument identifier inst, unit identifier u, subject identifier, protocol identifier, time point, and one or more measurement values. For parallel measurements, the mean, standard deviation, coefficient of variation, acquisition channel, and instrument quality control status for each measurement can also be recorded. The original field names and original measurement values ​​are always retained and are not overwritten by the standardized results.

[0029] In step S3, the set of measurable indicators for the instrument is read from the IMR based on the instrument identifier. The candidate standard indicator domain is preferentially formed by the intersection of the instrument's measurable indicator set and the program's target standard indicator set. When this intersection is empty, the system can use the instrument's measurable indicator set, the program's target standard indicator set, or the global standard indicator library as a fallback candidate domain, while applying penalty scores to candidates that do not belong to either of the aforementioned sets. This process restricts the field text to be interpreted only within a semantic space compatible with the current instrument capabilities and the current program objectives.

[0030] Subsequently, the system performs exact matching, prefix matching, restricted substring matching, and alias matching on x, from high to low determinism. Restricted substring matching only results in a candidate match when x and the candidate standard indicator name are mutually inclusive and their length difference is less than a preset threshold Δ; Δ can be configured by different indicator categories. This restriction is suitable for avoiding L value, a value, b Short strings such as values ​​may be mistakenly associated with each other under normal containment matching. Alias ​​matching can perform case correction, full-width / half-width character setting, space and symbol normalization before matching, but it does not change the original saved field name.

[0031] For each candidate criterion c, calculate the overall matching score. The overall matching score is derived from the text or alias matching score. Instrument compatibility score Unit compatibility score Scheme Context Score And numerical range or channel compatibility score At least three items are determined. Each scoring item is normalized to 0 to 1, with a preset non-negative weight and a weight sum of 1. Text scores are determined in descending order of priority: exact match of standard name, exact match of alias, prefix match, restricted substring match, and alias inclusion match.

[0032]

[0033] in, The comprehensive matching score for candidate standard indicator c; to The preset weights for the corresponding scoring items; , , , and These are scores for text, instrument, unit, scheme context, and numerical range or acquisition channel compatibility. Optionally, to The values ​​are set to 0.40, 0.20, 0.15, 0.15 and 0.10 respectively, but the specific values ​​are determined by the mapping configuration.

[0034] The highest score is recorded as S1, and the second highest score as S2. The system outputs a uniquely determined status and corresponding standard indicator only when S1 is not lower than the reception threshold θa and the difference between S1 and S2 is not lower than the interval threshold θm. If S1 is lower than θa, an unrecognized status is output; if S1 reaches θa but the difference is less than θm, a conflict status is output. Unrecognized or conflicting records save the original fields, instrument, unit, value, and candidate list, but by default, they are not included in the automatic efficacy determination.

[0035] Once unidentified or conflicting records are manually confirmed or a standard indicator is specified in the scheme-level mapping configuration, the system writes the relationship between the original fields, instrument identifiers, and unit identifiers to the standard indicator identifiers back to the mapping library and forms a new mapping version. This version can be reused when the same context occurs again in the same scheme. For records that are indeed non-standard test dimensions, a new standard indicator can be created and its unit, direction, threshold, or stability limit can be defined, and then incorporated into the required effectiveness through POR.

[0036] In step S4, the record that uniquely determines the state is written to the standard indicator identifier, mapping confidence, mapping state, candidate list, and mapping version. Optionally, mapping fingerprints are calculated for these fields and the participating candidate configurations. Step S5 only allows MJR retrieval using the standard indicator identifier as the key; therefore, once an original field is standardized to different standard indicators, it will obtain different directions of change, unit conversions, and decision paths.

[0037] The MJR (Mean Jurisprudential Jurisprudential Ratio) includes at least the direction, threshold, unit, and rule version. When the direction is increasing or decreasing, the indicator is directional; when the direction is stable, the indicator is stable. Rules may also include allowed instrument sets, allowed unit sets, unit conversion relationships, stability limit δ, minimum effective sample size, allowable missing rate, and data quality thresholds. If instruments or units are incompatible, the system blocks statistical merging, or first performs the unit and calibration conversions specified in the rules.

[0038] For directional indicators, the evidence at a given follow-up time point consists of the baseline mean, follow-up mean, statistical test results, relative rate of change, and effective sample size. The system outputs strong directional evidence, trend evidence, reverse evidence, or no valid evidence based on significance, actual direction of change, and minimum magnitude of change. For stable indicators, the system does not use significant increases or decreases as valid conclusions, but rather forms strong stable evidence, weak stable evidence, or unstable evidence based on stability limits or equivalent intervals.

[0039] Taking a stable indicator as an example, let the baseline mean be B, the follow-up mean be T, and the minimum reference value be ε, which is greater than 0. The rate of change is calculated as T relative to B. Strong stability evidence is formed when the confidence interval of the mean difference falls completely within the preset stability interval; weak stability evidence is formed when the rate of change falls within the stability limit but the confidence interval does not fall completely within the stability interval; and unstable evidence is formed in all other cases. ε is used to avoid unstable division calculations caused by a baseline mean of zero or close to zero.

[0040]

[0041] Where C is the relative rate of change, B is the baseline mean, T is the follow-up mean, and ε is the minimum reference value to prevent division by zero from a zero baseline. The stability limit δ is given by the MJR of the corresponding standard index and can be achieved using the results of relative change stability limit, absolute change stability limit, equivalence test, or non-inferiority test.

[0042] The mapping confidence level, along with the data quality derived from the effective sample size, missing rate, coefficient of variation for parallel measurements, and instrument quality control status, jointly participates in evidence gating. When the mapping confidence level or data quality falls below the automatic judgment threshold, at most trend evidence or weakly stable evidence is output at that time point, and strong evidence is not automatically output. If this gating is omitted, records with low-confidence mappings may be incorrectly included in strong power conclusions even when statistically significant, causing a disconnect between standardized uncertainty and the evaluation conclusions.

[0043] In step S6, the evidence status of the standard indicator m at time point t is converted into an evidence value. Strong directional or strong stable evidence is assigned a value of 1; trend evidence or weak stable evidence is assigned a value of 0.5; and reverse evidence, unstable evidence, or no valid evidence is assigned a value of 0. Weighting is based on time points. By integrating various time points, evidence for a single indicator across different time points can be obtained. The weights can be set based on the main visitor's viewpoint in the plan; if no weight is set, the same weight will be applied.

[0044]

[0045] in, For single-indicator cross-time point evidence of the standard indicator m, The weight of time point t, This represents the evidence value of indicator m at time point t. If there is no valid time point or the denominator is zero, the indicator outputs no valid evidence and no automatic strong conclusion is made.

[0046] For efficacy k, only standard indicators that are associated with efficacy k and have valid measurement data are summarized. Indicator weights are assigned to indicator m. And read the mapping confidence. The cross-indicator evidence for individual efficacy is calculated according to the following formula. Records in conflicting or unidentified states are not included in the automatic aggregation; gating conditions can also be set, such as the absence of strong reverse evidence for the main evaluation indicator or the main evaluation indicator meeting the minimum evidence requirements.

[0047]

[0048] in, For single-item efficacy evidence of efficacy k, The preset index weights for the standard index m. The mapping confidence level of the standard index m This is evidence for a single indicator across time points for the standard indicator m. When the denominator is zero, the system outputs "No Data" or "Manual Review Status," and does not replace it with the default score. The corresponding conclusion intervals and gating conditions are configured by the scheme, and each efficacy is calculated independently. The higher-level conclusion of any efficacy cannot cover other efficiencies.

[0049] In Example 1, a human efficacy evaluation scheme for moisturizing was established, including a baseline time point T0 and at least one follow-up time point. EMR correlates moisturizing with skin moisture content and transepidermal water loss (TEWL); the original imported fields are Corneometer and TEWL. Corneometer is standardized to skin moisture content after alias matching and validation by instrument, unit, and scheme context, and is bound to the direction of increase; TEWL is standardized to transepidermal water loss and bound to the direction of decrease. Subsequent modules generate time-point evidence based on their respective rules, and then generate a conclusion on moisturizing efficacy.

[0050] In Example 2, the colorimeter field L is imported. a b Alternatively, fields with additional descriptions may be provided. The system first performs exact and prefix matching, allowing restricted substring matching only if a match is missed. If the original field provides only short characters and the unit, instrument, or scheme context does not allow the candidate score to meet the interval threshold, the system outputs a conflict and does not automatically select the corresponding standard indicator.

[0051] In Example 3, a moisturizing solution uses a transdermal water loss tester. The original field is transdermal water loss, measured in units of mass per square meter per hour. The intersection of the instrument's measurable index set and the solution's target standard index set includes TEWL and skin moisture content. Since TEWL is compatible with the instrument, units, and aliases, while skin moisture content lacks unit and instrument compatibility, TEWL is uniquely determined after receiving the highest score and meeting the reception and interval thresholds.

[0052] In Example 4, a stability limit is set for the stable indicator, and the rule version is written into the time-point evidence. When the stability limit is subsequently updated, the historical report continues to reference its original mapping version and rule version; if recalculation with the new rule is required, a new chain of evidence version is established to coexist with the old results, without overwriting the conclusion path of the original report.

[0053] In step S7, the reporting module constructs a report context from the protocol, subjects, and statistically completed measurement records. Ordinary variable placeholders can be replaced with variable values, and dynamic content blocks can generate statistical tables and efficacy results based on the established evidence objects. The reporting module no longer interprets raw fields but only consumes standard indicators, evidence at each level, and version fingerprints, ensuring that different Word masters or other structured report formats do not alter the calculation results of the same chain of evidence.

[0054] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for evaluating the efficacy of cosmetics based on semantic mapping of efficacy, characterized in that, include: S1: Receive efficacy claim information from cosmetic human efficacy evaluation schemes, and determine the set of target standard indicators for the scheme based on the preset efficacy-standard indicator relationship and scheme-level coverage relationship. S2: Receive heterogeneous measurement records from cosmetic human efficacy evaluation instruments. The heterogeneous measurement records include at least the original field names, instrument identifier, unit identifier, subject identifier, protocol identifier, time point, and measurement value, and retain the original field names. S3: Obtain the set of measurable indicators of the instrument corresponding to the instrument identifier according to the instrument-indicator relationship, and construct a candidate standard indicator domain using the set of measurable indicators of the instrument and the set of target standard indicators of the scheme; S4: For each candidate standard indicator in the candidate standard indicator domain, perform multi-factor scoring based on the text or alias matching results between the original field name and the candidate standard indicator, instrument compatibility results, unit compatibility results, scheme context results, and numerical features or acquisition channel compatibility results, and determine the unique determination state, conflict state, or unidentified state according to the highest score reception threshold and the interval threshold between the highest score and the second highest score. S5: Only under the uniquely determined state, the corresponding standard indicator identifier, mapping confidence level and mapping version are written into the heterogeneous measurement record, and the standard indicator identifier is used to address the indicator judgment rules containing the expected change direction, change threshold, stability limit, unit compatibility condition and data quality condition. S6: Based on the indicator judgment rules, measurement results at each time point, mapping confidence and data quality, form time point evidence, summarize it into single indicator cross-time point evidence according to standard indicators, and summarize it into single efficacy cross-indicator evidence according to the standard indicators associated with efficacy claims. S7: Generate a traceable evidence chain, which is associated with the original measurement record, standard indicator identifier, mapping version, indicator judgment rule version, time point evidence, single indicator cross-time point evidence, single effect cross-indicator evidence, and report output content.

2. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: The efficacy-standard indicator relationship is used to record the correspondence between efficacy claims and the default standard indicator set. The scheme-level coverage relationship is used to replace, add, or delete the default standard indicator set of the current scheme. The scheme target standard indicator set also serves as a constraint condition for data entry, statistical object selection, and candidate standard indicator domain construction.

3. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: The text or alias matching is performed in the order of exact matching, prefix matching, restricted substring matching, and alias matching; among them, restricted substring matching is performed only when the original field name and the candidate standard indicator name are mutually inclusive and the length difference between them is less than a preset length difference threshold.

4. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: When the intersection of the instrument measurable index set and the scheme target standard index set is not empty, the candidate standard index domain is the intersection; when the intersection is empty, at least one of the instrument measurable index set, the scheme target standard index set, or the global standard index set is used as the backoff candidate domain, and a scoring penalty is applied to candidate standard indices that do not belong to the instrument measurable index set or the scheme target standard index set.

5. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: The multi-factor scoring includes at least three of the following: text matching score, instrument compatibility score, unit compatibility score, scheme context score, and numerical range or acquisition channel compatibility score. When the highest score is lower than the highest score reception threshold, it is determined to be in an unrecognized state. When the highest score is not lower than the highest score reception threshold and the difference between the highest score and the second highest score is lower than the interval threshold, it is determined to be in a conflict state.

6. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: For directional indicators that change in the direction of increase or decrease, strong directional evidence, trend evidence, reverse evidence, or no valid evidence are formed based on statistical significance, actual direction of change, and minimum magnitude of change. For stable indicators that change in the direction of stability, strong stable evidence, weak stable evidence, or unstable evidence are formed based on whether the change of the follow-up value relative to the baseline value meets the preset stability limit or equivalent interval.

7. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 6, characterized in that: When the mapping confidence or data quality is lower than the preset automatic judgment threshold, at most trend evidence or weakly stable evidence will be formed at the corresponding time point, but no strong directional evidence or strongly stable evidence will be formed; the data quality is determined based on at least one of the following: effective sample size, missing rate, parallel measurement coefficient of variation, and instrument quality control status.

8. The cosmetic efficacy evaluation method based on efficacy semantic mapping according to claim 1, characterized in that: Evidence values ​​of a single indicator at each time point are weighted and fused according to the time point weight to obtain cross-time point evidence for a single indicator; cross-time point evidence of a single indicator corresponding to the standard indicator associated with the same efficacy claim is weighted and fused according to the indicator weight and mapping confidence to obtain cross-indicator evidence for a single efficacy claim; heterogeneous measurement records with a mapping status of conflict or unidentified status are not included in the automatic aggregation.

9. A cosmetic efficacy evaluation system based on efficacy semantic mapping, applied to the cosmetic efficacy evaluation method based on efficacy semantic mapping according to any one of claims 1-8, characterized in that, It includes a scheme semantic module, a record receiving module, a candidate domain construction module, a conflict resolution module, a rule addressing module, an evidence summarization module, and an evidence chain output module; The scheme semantic module is used to determine the set of target standard indicators for the scheme; the record receiving module is used to receive and retain the original field names of heterogeneous measurement records and their instrument, unit, and time point context; the candidate domain construction module is used to construct candidate standard indicator domains based on the set of measurable indicators of the instrument and the set of target standard indicators for the scheme; the conflict resolution module is used to output a uniquely determined state, a conflict state, or an unidentified state based on multi-factor scoring and receiving thresholds and interval thresholds; the rule addressing module is used to write standard indicator identifiers, mapping confidence levels, and mapping versions only for records with uniquely determined states, and uses standard indicator identifiers to address indicator judgment rules; the evidence summarization module is used to form time point evidence, single indicator cross-time point evidence, and single-item efficacy cross-indicator evidence; the evidence chain output module is used to associate the original measurement records, mapping versions, rule versions, evidence results, and report output content.

Images