Real estate project value prediction method and system
By combining the entropy weight method and the Gaussian kernel function, the instability problem of commercial real estate rental forecasting under small sample size is solved, and robust value forecasting is achieved in commercial and retail real estate rental forecasting.
Patent Information
- Application Number
- CN202610368834.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-24
- Publication Date
- 2026-05-15
AI Technical Summary
Existing technologies, especially in the case of small sample sizes, are unable to effectively combine the key differences between the product and its competitors in commercial and retail real estate rental forecasting. This results in unstable inference results that are easily affected by extreme values, making it impossible to accurately predict the rental level of the product's project.
The entropy weight method is used to process the feature data, calculate the scores of the product and competitors, and combine the Gaussian kernel function and linear extrapolation method to perform weighted fusion and correction to ensure the robustness and rationality of the prediction results.
With a small sample size, it can accurately predict the rental level of this project. By combining the key differences between this product and competitors, the uncertainty is quantified, thereby improving the rationality and credibility of the reasoning results.
Smart Images
Figure CN122048398A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a method and system for predicting the value of real estate projects. Background Technology
[0002] Rental levels for commercial and retail real estate are influenced by a combination of factors, including commercial building area, number of brands, foot traffic density, size of the surrounding resident population, number of office workers, and prices of secondhand homes in the vicinity.
[0003] Common valuation and inference methods generally fall into two categories. One is the similarity method, which estimates the target rent by combining the similarity between the product's and competitors' data and using a weighted average of competitors' rents. Its advantage lies in its intuitive logic and easily interpretable results. However, in practical applications, when the number of available competitor samples is limited or the product itself has a significant advantage, this method struggles to provide reasonable extrapolation estimates. The other category employs regression analysis or machine learning algorithms. These methods rely on a large amount of historical sample data to train the predictive model. Their main advantage is their strong predictive potential, capable of capturing complex relationships. However, when the number of training samples is insufficient, the model is prone to overfitting, thus affecting the reliability of the results. Summary of the Invention
[0004] The embodiments of this application aim to at least partially address one of the technical problems in the related art. To this end, the embodiments of this application propose a method and system for predicting the value of real estate projects.
[0005] Embodiments of this application provide a method for predicting the value of a real estate project. The method includes: acquiring feature data of the project itself and feature data of competitors; preprocessing the feature data of the project itself and feature data of competitors to obtain numerical feature values and normalized feature values; calculating the project score of the project itself and the project score of competitors using the entropy weight method based on the numerical feature values and normalized feature values; predicting the integrated value data of the project itself based on the project score of the project itself and the project score of competitors; and correcting the integrated value data of the project itself based on the numerical feature values to obtain the target value prediction result of the project itself.
[0006] In some embodiments, based on the product project score and the competitor project score, the fusion value data of the product project is predicted, including: calculating the first predicted value data of the product project using a Gaussian kernel function based on the product project score and the competitor project score, and calculating the second predicted value data of the product project using linear extrapolation; determining the fusion weight according to the score interval relationship between the product project score and the competitor project score; and performing weighted fusion of the first predicted value data and the second predicted value data based on the fusion weight to obtain the fusion value data of the product project.
[0007] In some embodiments, based on the product project score and the competitor project score, a first predicted value data for the product project is calculated using a Gaussian kernel function, and a second predicted value data for the product project is calculated using linear extrapolation. This includes: calculating the absolute distance between the product project score and the competitor project score; normalizing the absolute distance to obtain a normalized distance; calculating the competitor project value weight based on the normalized distance and the Gaussian kernel function; performing a weighted summation of the competitor project values based on the competitor project value weight to obtain the first predicted value data; sorting the competitor projects according to their scores, and comparing the product project score with the sorted competitor project scores to determine the second predicted value data.
[0008] In some embodiments, competitor items are sorted according to their scores, and the score of the product item is compared with the sorted competitor item scores to determine a second predicted value data. This includes: sorting competitor items in ascending order according to their scores to obtain a competitor item score range, wherein the competitor item score range includes a maximum value and a minimum value; calculating a sensitivity slope based on competitor item scores and competitor item values with a preset positional relationship within the competitor item score range; determining the position of the product item score within the competitor item score range, and determining the second predicted value data based on different positional relationships.
[0009] In some embodiments, determining the position of the product's score within the range of competitor's scores, and determining the second predicted value data based on different positional relationships, includes: if the product's score is higher than the maximum value of the range, then linearly extrapolating to the first side based on the sensitivity slope, the competitor's score corresponding to the maximum value of the range, and the competitor's value, to obtain the second predicted value data, wherein the first side represents the side where the score is higher than the maximum value of the range; If the score of this product is lower than the minimum value of the interval, then based on the sensitivity slope, the score of the competitor's product corresponding to the minimum value of the interval, and the value of the competitor's product, a linear extrapolation is performed to the second side to obtain the second predicted value data, where the second side represents the side with a score lower than the minimum value of the interval. If the score of this product falls within the score range of competing products, then the score of this product is linearly interpolated based on the ranked scores and values of competing products to obtain the second predicted value data.
[0010] In some embodiments, the product feature data and competitor feature data are preprocessed to obtain numerical feature values and normalized feature values, including: performing numerical mapping transformation on the text features in the product feature data and competitor feature data to obtain numerical feature values; and based on preset positive and negative indicators, normalizing the numerical feature values corresponding to the positive indicators and inverting and normalizing the numerical feature values corresponding to the negative indicators to obtain normalized feature values.
[0011] In some embodiments, the product score and competitor score are calculated using the entropy weight method based on numerical and normalized feature values. This includes: calculating the information entropy value corresponding to each feature using the entropy weight method based on the numerical and normalized feature values to obtain the feature information entropy; calculating the feature weight based on the feature information entropy; and calculating the product score and competitor score by weighted summation based on the feature weight and the corresponding normalized feature value. In some embodiments, the numerical feature values include the feature values of the current product project and the feature values of competing products projects. The process of correcting the integrated value data of the current product project based on the numerical feature values to obtain a target value prediction result for the current product project includes: calculating the mean ratio between the feature values of the current product project and the feature values of competing products projects based on key features selected from the feature values of the current product project and the feature values of competing products projects, and using the mean ratio as the feature advantage ratio; extrapolating and correcting the integrated value data of the current product project based on the feature advantage ratio and a preset correction coefficient to obtain the corrected value of the current product project; and applying constraints to the corrected value of the current product project to obtain a target value prediction result for the current product project. In some embodiments, the modified value of the product project is constrained to obtain the target value prediction result of the product project, including: calculating the upper limit and lower limit of the constraint value of the product project based on the maximum value, minimum value and sample size of the competitor project; constraining the modified value of the product project between the upper limit and lower limit of the constraint value to obtain the target value prediction result.
[0012] Embodiments of this application provide a real estate project value prediction system. The system includes: an acquisition module for acquiring feature data of the project itself and feature data of competitors; a processing module for preprocessing the feature data of the project itself and feature data of competitors to obtain numerical feature values and normalized feature values; a calculation module for calculating the project score of the project itself and the project score of competitors based on the numerical feature values and normalized feature values using the entropy weight method; a prediction module for predicting the integrated value data of the project itself based on the project score of the project itself and the project score of competitors; and a correction module for correcting the integrated value data of the project itself based on the numerical feature values to obtain the target value prediction result of the project itself.
[0013] Embodiments of this application provide an electronic device, which includes: a memory, and one or more processors communicatively connected to the memory; the memory stores instructions executable by the one or more processors, which are executed by the one or more processors to cause the one or more processors to implement the steps of the method of any of the above embodiments.
[0014] Embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method of any of the above embodiments.
[0015] Embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the method according to any of the above embodiments.
[0016] The solution provided in this application involves acquiring and preprocessing the product's feature data and competitors' feature data, converting them into numerical and normalized feature values. Then, the entropy weight method is used to calculate the product's project score and competitors' project score. Based on these scores, the integrated value data of the product's project is predicted. Finally, the integrated value data of the product's project is specifically corrected based on the numerical feature values, thereby obtaining the target value prediction result for the product's project. It is evident that the solution provided in this application, even with a small sample size, can both incorporate the key differences between the product and competitors and quantify uncertainty, thereby improving the rationality and credibility of the reasoning results and achieving robust reasoning to obtain the target value prediction result for the product's project. Attached Figure Description
[0017] Figure 1 A flowchart illustrating a method for predicting the value of a real estate project, provided as an embodiment of this application; Figure 2 A flowchart illustrating the real estate project value prediction method provided in this application embodiment; Figure 3 A schematic diagram of a real estate project value prediction system provided in this application embodiment; Figure 4 A block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation
[0018] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.
[0019] In commercial real estate or shop leasing scenarios, it's often necessary to infer a reasonable rental level for a given project based on limited data from competing projects. Some examples employ regression analysis or machine learning algorithms; however, many traditional machine learning or statistical methods are highly dependent on sample size, typically requiring large datasets to achieve a stable fit. In real-world projects, the number of comparable competitors available is often very limited, sometimes even just a few, making it difficult for these methods to achieve their intended effectiveness.
[0020] Some examples of inference methods based on similarity or mean often rely solely on the average rent of competitors to infer the rental level of the project itself, failing to effectively incorporate the advantage ratio, i.e., the comparison of deviations between key indicators, thus making the inference results prone to large biases.
[0021] When there are extreme values or significant differences in project characteristics among the competing samples in some examples, the inference process is prone to over-correction, which brings obvious extrapolation risks and makes the prediction results highly unstable.
[0022] Some examples calculate cosine similarity based on the similarity features between the product and competing products. After normalizing the similarity scores, they are weighted and summed with the average rent of competing products to obtain the average rent of the product.
[0023] However, in traditional reasoning methods, the rental income of this project often cannot exceed the maximum rental income of competing projects. When this project outperforms competing projects in key indicators such as weekday or holiday foot traffic, or in the average price details of second-hand homes within a five-kilometer radius of this project, the maximum rental income of this project is still limited to the maximum rental income of competing projects because the reasoning process relies on similarity logic. Therefore, it is impossible to obtain a reasoning result that exceeds the maximum rental income of competing projects.
[0024] Meanwhile, due to the small sample size, a project of this product usually only has 3 to 5 competitor data. In this case, the weight allocation and the rental value of a single competitor are easily affected by extreme values, which leads to the overall inference result being dominated by outliers and reducing the accuracy of the inference result.
[0025] In view of the above, this application provides a method for predicting the value of real estate projects, which can ensure robust inference of the project's rental income with a small sample size. It can combine the key differences between the project and its competitors, and quantify the uncertainty, thereby improving the rationality and credibility of the inference results.
[0026] Figure 1 This is a flowchart illustrating a method for predicting the value of a real estate project, as provided in an embodiment of this application.
[0027] like Figure 1As shown, the real estate project value prediction method 100 provided in this application embodiment includes steps S110-S150.
[0028] Step S110: Obtain the product feature data and competitor product feature data.
[0029] For example, a real estate project to be predicted can be selected as the product project, and feature data of multiple competing projects comparable to the product project can be obtained. These feature data may include rent-related features, regional business district features, and commercial building area features, etc.
[0030] Step S120: Preprocess the feature data of this product and the feature data of competitors to obtain numerical feature values and normalized feature values.
[0031] For example, the system can perform missing value imputation on the acquired feature data, filling in missing data in each feature field using an appropriate method. The system can also perform numerical conversion on text-type features, transforming text features into numerical representations. Then, the system normalizes the numerical feature values, mapping them to a uniform numerical range to obtain normalized feature values.
[0032] Step S130: Based on the numerical and normalized feature values, the entropy weight method is used to calculate the score of this product and the score of the competitor's product.
[0033] For example, the system can calculate the weight of each feature using the information entropy formula and assign a corresponding weight value to each feature. Then, the system combines the normalized value of the feature and the weight of the feature to calculate the score of the sample, thus obtaining the score of the product and the score of the competitor.
[0034] Step S140: Based on the product's project score and the competitor's project score, predict the product's integrated value data.
[0035] For example, the system performs rental inference based on the scores of the product project sample and the score range of the competitor project sample, and predicts the integration value data of the product project.
[0036] Step S150: Based on the numerical feature values, the integrated value data of this product project is corrected to obtain the target value prediction result of this product project.
[0037] For example, the system can calculate the advantage ratio based on numerical feature values and according to the advantage ratio calculation formula, and then combine the integrated value data of the project and the advantage ratio to correct the predicted rent of the project, thereby obtaining the target value prediction result of the project.
[0038] The solution provided in this application involves acquiring and preprocessing the product's feature data and competitors' feature data, converting them into numerical and normalized feature values. Then, the entropy weight method is used to calculate the product's project score and competitors' project score. Based on these scores, the integrated value data of the product's project is predicted. Finally, the integrated value data of the product's project is specifically corrected based on the numerical feature values, thereby obtaining the target value prediction result for the product's project. It is evident that the solution provided in this application, even with a small sample size, can both incorporate the key differences between the product and competitors and quantify uncertainty, thereby improving the rationality and credibility of the reasoning results and achieving robust reasoning to obtain the target value prediction result for the product's project.
[0039] Figure 2 This is a flowchart illustrating the method for predicting the value of real estate projects provided in this application embodiment.
[0040] like Figure 2 As shown, the real estate project value prediction method first performs data preprocessing, then uses the entropy weight method to calculate feature weights, then calculates sample scores, then performs distance interpolation baseline, then performs linear extrapolation of the project rent, then performs project inference rent fusion, then calculates the advantage ratio of key indicators, then performs extrapolation correction, then determines the upper and lower limits of the project rent, then outputs the project rent and explanatory indicators, and finally makes corrections based on business feedback.
[0041] The following will combine Figure 1 , Figure 2 The content of this application provides a detailed description of the solution provided.
[0042] In some embodiments of this application, the product feature data and competitor feature data are preprocessed to obtain numerical feature values and normalized feature values, including: performing numerical mapping transformation on the text features in the product feature data and competitor feature data to obtain numerical feature values; based on preset positive and negative indicators, normalizing the numerical feature values corresponding to the positive indicators, and inverting and normalizing the numerical feature values corresponding to the negative indicators to obtain normalized feature values.
[0043] For example, the system first determines the positive and negative indicators required for calculating the advantage ratio. Positive indicators refer to features that have a positive impact on rent, meaning the higher the indicator value, the more advantageous it is. These include daily average foot traffic density, daily average foot traffic density during holidays, the number of permanent residents within 5KM, the number of office workers, the average transaction price of second-hand homes within 5KM, and the number of brands. Negative indicators refer to features that have a negative impact on rent, meaning the lower the indicator value, the more advantageous it is. These include the existing commercial space within 5KM (in ten thousand square meters), the number of existing commercial properties within 5KM, the number of planned new commercial properties within 5KM, and the planned new commercial space within 5KM (in ten thousand square meters). A larger existing commercial space within 5KM indicates more intense competition and greater rent pressure; a larger number of existing and planned new commercial properties within 5KM indicates more commercial properties and a significant diversion effect; a larger planned new commercial space within 5KM indicates more planned projects and stronger expected competition.
[0044] Then, the system performs the conversion according to preset mapping rules, such as {"City-level business district":3,"Regional business district":2,"-":1}. This converts text-based regional business district features into numerical features: city-level business districts are mapped to 3, regional business districts to 2, and unlabeled or uninformed business districts are mapped to 1. The system also performs numerical processing on commercial building area features. For example, the system multiplies the area unit (in ten thousand square meters) by 10,000 to convert it into a numerical feature; for instance, a commercial building area of 50,000 square meters will be converted to 50,000.
[0045] Furthermore, the system also performs splitting processing on proportional features. For example, the system splits proportional features such as the gender ratio of the resident population, the family structure ratio of the resident population (young customers / young families / mature families), and the affluence ratio of the resident population (salaried / comfortable / middle class / wealthy) into multiple features by commas. For example, the original features might be 13.69%, 56.65%, 27.99%, and 1.67%. The system converts commercial floor features into the total number of floors. For example, floor information containing negative floors is converted into a total number of floors. For example, "-1 to 6 floors" is converted into a total of 7 floors. The system converts the opening date into the difference between the opening year and the current year.
[0046] After completing the above numerical processing, the numerical feature values can be normalized based on preset positive and negative indices. Specifically, the numerical features corresponding to positive indices can be directly normalized, while the numerical features corresponding to negative indices can be inverted before normalization.
[0047] In some embodiments of this application, the product score and competitor score are calculated using the entropy weight method based on numerical and normalized feature values. This includes: calculating the information entropy value corresponding to each feature using the entropy weight method based on numerical and normalized feature values to obtain the feature information entropy; calculating the feature weight based on the feature information entropy; and calculating the product score and competitor score by weighted summation based on the feature weight and the corresponding normalized feature value.
[0048] For example, after data preprocessing and obtaining normalized feature values, the entropy weight method can be used to calculate the importance of each feature, thereby determining the weight of each feature. The entropy weight method assigns weights based on the degree of difference between different items for each feature indicator. When a feature differs little (or is almost the same) between different items, its information content is low, its corresponding information entropy is high, and its assigned weight is low; when a feature differs greatly between different items, it can effectively distinguish between different items, its corresponding information entropy is low, and its assigned weight is high.
[0049] The weight of each feature is calculated using the information entropy formula. Specifically, each column of values is converted into a probability distribution: p = each value / the sum of the column, where p is the relative distribution of the index values, which facilitates the calculation of entropy.
[0050] Calculate entropy using the formula for entropy:
[0051] The entropy value ranges from 0 to 1; the closer it is to 1, the more uniform the distribution of the indicator (low discrimination); the closer it is to 0, the greater the variability of the indicator (high discrimination). Let m represent the entropy value of the j-th feature, and m represent the number of samples. This represents the normalized value of the i-th sample under the j-th feature. This represents the normalization constant.
[0052] After obtaining the information entropy of each feature, the dissimilarity of each feature can be further calculated. For example, the dissimilarity can be calculated using the formula d = 1.0. The variance values of each feature are calculated using `ej`, where a larger `d` indicates stronger discriminative power and greater importance. Then, the corresponding feature weights can be calculated based on the variance of each feature. For example, when the variance of all features is 0, a uniform distribution method can be used to assign weights to each feature, i.e., each feature has a weight of 1 / n, allocated according to the proportion of variance `d`: `weights = d / d.sum()`, where `n` represents the number of features; `d` represents the variance of the feature. The sum of the weights is guaranteed to be 1.
[0053] After obtaining the weights of each feature, the sample score can be calculated by combining the normalized feature values corresponding to each sample. For example, the score of each sample = the normalized value of each indicator × the weighted sum of the corresponding weights. The higher the score, the better the performance of the item under the weighted indicators.
[0054] In the embodiments provided in this application, the scores calculated using the entropy weight method fully consider the differences in characteristics, effectively distinguishing the value levels of different items. The scores for this product can be used for subsequent value prediction and correction, while the scores for competing products can be used to construct a comparison benchmark and calculate subsequent fused value data.
[0055] In some embodiments of this application, the fusion value data of the product project is predicted based on the product project score and the competitor project score, including: calculating the first predicted value data of the product project using a Gaussian kernel function based on the product project score and the competitor project score, and calculating the second predicted value data of the product project using linear extrapolation; determining the fusion weight according to the score interval relationship between the product project score and the competitor project score; and performing weighted fusion of the first predicted value data and the second predicted value data based on the fusion weight to obtain the fusion value data of the product project. In other embodiments, based on the product's score and the competitor's score, a first predicted value data for the product is calculated using a Gaussian kernel function, and a second predicted value data for the product is calculated using linear extrapolation. This includes: calculating the absolute distance between the product's score and the competitor's score; normalizing the absolute distance to obtain a normalized distance; calculating the competitor's value weight based on the normalized distance and the Gaussian kernel function; performing a weighted summation of the competitor's value based on the competitor's value weight to obtain the first predicted value data; sorting the competitor's projects according to their scores, and comparing the product's score with the sorted competitor's scores to determine the second predicted value data.
[0056] For example, the system calculates the absolute distance between the project's score and that of each competing project. Competitors with closer scores contribute more to the predicted rent. The system then normalizes the calculated absolute distances to ensure stability. For instance, normalization is performed using the formula d_norm = d / maxd, where d is the absolute distance, maxd is the maximum distance among all distances, and d_norm is the normalized distance. Each distance value is divided by the maximum distance to ensure that data of different magnitudes do not affect the Gaussian weight calculation.
[0057] The system calculates weights based on a Gaussian kernel (exponential decay), with closer distances resulting in larger weights. The system calculates weights using the formula w = exp(-gamma × d_norm), where gamma is a hyperparameter controlling the kernel width and d_norm is the normalized distance. Closer distances result in smaller exponential decay and larger weights. A larger gamma leads to faster decay, approximating the nearest neighbor; a smaller gamma results in a more even distribution of influence across competing products.
[0058] The system normalizes the calculated weights so that the sum of the weights is 1. The system normalizes according to the formula w = w / w.sum(), where w represents the weights calculated in the previous step, and w.sum() represents the sum of all weights. The system then performs a weighted summation of the competitor's rental prices to obtain the first predicted value data for this project.
[0059] In other embodiments, competitor items are sorted according to their scores, and the score of the product item is compared with the scores of the sorted competitor items to determine the second predicted value data. This includes: sorting competitor items in ascending order according to their scores to obtain a competitor item score range, wherein the competitor item score range includes a maximum value and a minimum value; calculating a sensitivity slope based on the competitor item scores and competitor item values with a preset positional relationship within the competitor item score range; determining the position of the product item score within the competitor item score range, and determining the second predicted value data based on different positional relationships.
[0060] In other embodiments, determining the position of the product's score within the competitor's score range, and determining second predicted value data based on different positional relationships, includes: if the product's score is higher than the maximum value of the range, then linearly extrapolating to a first side based on the sensitivity slope, the competitor's score corresponding to the maximum value of the range, and the competitor's value, to obtain second predicted value data, wherein the first side represents the side where the score is higher than the maximum value of the range; if the product's score is lower than the minimum value of the range, then linearly extrapolating to a second side based on the sensitivity slope, the competitor's score corresponding to the minimum value of the range, and the competitor's value, to obtain second predicted value data, wherein the second side represents the side where the score is lower than the minimum value of the range; if the product's score is within the competitor's score range, then linearly interpolating the product's score based on the sorted competitor's scores and competitor's values to obtain second predicted value data.
[0061] The system sorts the data in ascending order based on competitor scores. If the number of competitor samples is less than 2, it degenerates into a simple average as the prediction; otherwise, it uses the first two sample points to calculate the interval slope. The system calculates the slope of the first interval using the following formula:
[0062] in, This represents the difference between the rent for competitor 2 and the rent for competitor 1. This represents the score of competitor 2 minus the score of competitor 1. The slope represents the "score's sensitivity to rent," i.e., the sensitivity slope. Extrapolation is the prediction extended along this slope.
[0063] The sorted score sequence s_sorted and the corresponding rent sequence r_sorted are obtained.
[0064] The system calculates the second predicted value data by selecting different extrapolation rules based on the positional relationship between the score of the product item and the score interval of the competitor item. Specifically, when the score of the product item is higher than the score of the maximum competitor item, the system extrapolates along the slope based on the maximum score point. The system calculates according to the formula delta_score = score_target - s_sorted[-1] and rent_pred = r_sorted[-1] + slope × delta_score, where score_target represents the score of the product item sample of the rent to be inferred, s_sorted[-1] represents the score of the sample with the highest score of the competitor item, delta_score represents the score difference between the two, slope represents the slope obtained above, and r_sorted[-1] represents the rent corresponding to the sample with the highest score of the competitor item.
[0065] When the score of the product item is lower than the minimum score of the competitor item, the minimum score point is used as the benchmark, and extrapolation is performed along the slope. The system calculates according to the formula delta_score = score_target - s_sorted[0] and rent_pred = r_sorted[0] + slope × delta_score, where score_target represents the sample score of the product item to be inferred rent, s_sorted[0] represents the sample score of the competitor item with the lowest score, delta_score represents the score difference between the two, slope represents the slope obtained in the previous step, and r_sorted[0] represents the rent corresponding to the sample with the lowest competitor item score.
[0066] When the score of this product falls within the score range of competitors, linear interpolation is used. The system calculates rent_pred according to the formula rent_pred = np.interp(score_target, s_sorted, r_sorted), where score_target represents the sample score of this product, s_sorted is the score sequence of competitors, r_sorted is the rent sequence of competitors, np represents the NumPy library in Python, and np.interp is the linear interpolation function.
[0067] The system uses the rent_pred calculated above as the second predicted value data for this product project, which can be denoted as rent_slope.
[0068] The system obtains the maximum value (smax) and minimum value (smin) of the product's score and the competitor's sample scores, determining the competitor's score range as [smin, smax]. It then determines the position of the product's score within the competitor's score range. If the product's score falls within the competitor's score range, the rental inference for the product's project primarily uses distance interpolation, with a fusion weight of w_interp of 0.8 and w_extrap of 0.2.
[0069] When the score of this product item is outside the score range of the competitor's product item, the standardized distance norm_out between this product item and the competitor's score range boundary is calculated based on the extrapolated rent. If the score of this product item is lower than the minimum value of the competitor's score range, then norm_out is calculated according to the formula norm_out = (smin - score_target) / (smax - smin + 1e-12), where score_target is the score of this product item, and smax and smin represent the maximum and minimum values of the competitor's score range, respectively.
[0070] If the score of this product is higher than the maximum score range of the competitors, then the formula norm_out = (score_target - smax) / (smax - smin + 1e-12) is used for calculation. The system maps norm_out to the range [0.05, 0.95] to control the range of interpolation weights, resulting in the final fusion weights w_interp and w_extrap.
[0071] Finally, the system performs a fusion calculation according to the formula rent_combined_raw = w_interp * rent_dist + w_extrap * rent_slope, where rent_dist is the first predicted value data (distance interpolation prediction), rent_slope is the second predicted value data (extrapolation prediction), w_interp is the interpolation weight, and w_extrap is the extrapolation weight. The system calculates the final raw prediction as: Distance Interpolation Prediction × Interpolation Weight + Extrapolation Prediction × Extrapolation Weight, thus obtaining the fusion value data for this project.
[0072] In some embodiments of this application, the numerical feature values include the feature values of the product project and the feature values of competing products. The process of correcting the integrated value data of the product project based on these numerical feature values to obtain a target value prediction result for the product project includes: calculating the mean ratio between the feature values of the product project and the feature values of competing products based on key features selected from the feature values of the product project and the feature values of competing products, and using this mean ratio as the feature advantage ratio; extrapolating and correcting the integrated value data of the product project based on the feature advantage ratio and a preset correction coefficient to obtain a corrected value for the product project; and constraining the corrected value of the product project to obtain a target value prediction result for the product project.
[0073] For example, the system selects key features from the feature values of the product and those of competitors, calculates the mean ratio between the feature values of the product and those of competitors, and uses the mean ratio as the feature advantage ratio.
[0074] The advantage ratio for each feature can be logarithmically transformed, and the average of the transformed results can be calculated. Then, an exponential operation is used to obtain the geometric mean, thus yielding the overall advantage ratio. This method reduces the impact of extreme values on the calculation results. When the geometrically averaged advantage ratio is greater than 1, it indicates that this product / project is generally superior to competitors in key indicators; when the advantage ratio is less than 1, it indicates that this product / project is generally inferior to competitors in key indicators.
[0075] The system combines the integrated value data of this product project with the advantage ratio calculated above, and adjusts the rent of this product project according to the formula final_rent = base_rent × (1 + α × (advantage_ratio - 1)). Here, base_rent is the integrated value data of this product project, advantage_ratio is the geometric mean advantage ratio of this product project relative to competitors, and α is a correction coefficient that controls the amplification or reduction of the advantage ratio when it deviates from 1.
[0076] When the eigenvalue ratio is greater than 1, it indicates that the project is superior to competitors in key indicators, and the correction coefficient α is positive, increasing the predictive value of the project through extrapolation. When the eigenvalue ratio is less than 1, it indicates that the project is inferior to competitors, and the correction coefficient α is positive, decreasing the predictive value of the project through extrapolation. When the eigenvalue ratio is equal to 1, it indicates that the project is on par with competitors, and no correction is made. The system uses the corrected value as the corrected value of the project.
[0077] In some embodiments of this application, the modified value of the product project is constrained to obtain the target value prediction result of the product project, including: calculating the upper limit and lower limit of the constraint value of the product project based on the maximum value, minimum value and sample size of the competing products project; constraining the modified value of the product project between the upper limit and lower limit of the constraint value to obtain the target value prediction result.
[0078] For example, the system can obtain the maximum value comp_max and minimum value comp_min of the competitor's project, as well as the sample number n_comp of the competitor's project rent, and calculate the range span of the competitor's value span = comp_max - comp_min.
[0079] The system calculates the upper bound coefficient cap_extrapolation(n) based on the number of competitor project samples n_comp. A larger sample size results in a smaller extrapolation range, while a smaller sample size results in a larger extrapolation range. The system then calculates the lower and upper bounds of the project's value constraint according to the formulas lower = comp_min - cap_extrapolation(n) × span and upper = comp_max + cap_extrapolation(n) × span. Here, lower represents the lower bound of the project's value constraint, upper represents the upper bound, comp_min represents the minimum rent among competitor projects, comp_max represents the maximum rent among competitor projects, and n_comp represents the number of sample rents for competitor projects.
[0080] The final output of the system for the project's rental value equals the extrapolated adjusted value combined with the predicted value after truncation by upper and lower limits. If the adjusted value of the project is greater than the upper limit, the upper limit is used; if the adjusted value is less than the lower limit, the lower limit is used; if the adjusted value is within the constraints, the adjusted value itself is used. Through constraint truncation, the system ensures that the predicted target value of the project is within a reasonable range of competing product values.
[0081] The system outputs the final target value prediction result for this project and explanatory indicators used in the reasoning process. The target value prediction result includes, for example, the project rent after constraint truncation. Explanatory indicators include, for example, feature weights using the entropy weight method, sample scores, the project rent calculated using a Gaussian kernel, the project rent extrapolated linearly, dominant features, advantage ratio, constraint range, and the final project score. The system can also incorporate business feedback to adjust other hyperparameters in the calculated project rent, such as the correction coefficient alpha, to control the amplification or reduction of the advantage ratio when it deviates from 1.
[0082] As can be seen, the embodiments of this application provide a hybrid inference method that combines entropy weighting with similarity, extrapolation correction, and small-sample robustness constraints. By combining information entropy to calculate the difference weight score for each feature, and then calculating the similarity-weighted average of competitor rental prices based on score distance, and the weighted average of the linearly extrapolated rental prices as the rental price of the current product, the upper and lower limits of the inferred rental price are calculated based on the rental range of competitors, the inferred rental price is corrected, and thus the target value prediction result of the current product is obtained.
[0083] Figure 3 This is a schematic diagram of a real estate project value prediction system provided in an embodiment of this application.
[0084] Embodiments of this application provide a real estate project value prediction system 300, comprising: The acquisition module 310 is used to acquire the feature data of this product and the feature data of competitors.
[0085] The processing module 320 is used to preprocess the feature data of this product and the feature data of competitors to obtain numerical feature values and normalized feature values.
[0086] The calculation module 330 is used to calculate the scores of the product and the competitor's products based on numerical and normalized feature values using the entropy weight method.
[0087] The prediction module 340 is used to predict the integrated value data of the product project based on the product project score and the competitor project score.
[0088] The correction module 350 is used to correct the integrated value data of the product project based on the numerical feature values to obtain the target value prediction result of the product project.
[0089] In some embodiments, the prediction module 340 is further configured to: calculate the first predicted value data of the product project using a Gaussian kernel function based on the product project score and the competitor project score, and calculate the second predicted value data of the product project using linear extrapolation; determine the fusion weight according to the score interval relationship between the product project score and the competitor project score; and perform weighted fusion of the first predicted value data and the second predicted value data based on the fusion weight to obtain the product project fusion value data.
[0090] In some embodiments, the prediction module 340 is further configured to: calculate the absolute distance between the scores of the product project and the competitor project based on the product project score and the competitor project score; normalize the absolute distance to obtain a normalized distance; calculate the competitor project value weight based on the normalized distance and the Gaussian kernel function; perform a weighted summation of the competitor project value based on the competitor project value weight to obtain first predicted value data; sort the competitor projects according to their scores, and compare the product project score with the sorted competitor project scores to determine second predicted value data.
[0091] In some embodiments, the prediction module 340 is further configured to: sort competitor items in ascending order according to competitor item scores to obtain competitor item score intervals, wherein the competitor item score interval includes a maximum value and a minimum value; calculate a sensitivity slope based on competitor item scores and competitor item values whose score relationship within the competitor item score interval is a preset positional relationship; determine the position of the product item score within the competitor item score interval, and determine a second predicted value data based on different positional relationships.
[0092] In some embodiments, the prediction module 340 is further configured to: if the score of the product item is higher than the maximum value of the interval, then perform linear extrapolation to the first side based on the sensitivity slope, the competitor item score corresponding to the maximum value of the interval, and the competitor item value to obtain second predicted value data, wherein the first side represents the side where the score is higher than the maximum value of the interval; if the score of the product item is lower than the minimum value of the interval, then perform linear extrapolation to the second side based on the sensitivity slope, the competitor item score corresponding to the minimum value of the interval, and the competitor item value to obtain second predicted value data, wherein the second side represents the side where the score is lower than the minimum value of the interval; if the score of the product item is within the range of competitor item scores, then perform linear interpolation on the score of the product item based on the sorted competitor item scores and competitor item values to obtain second predicted value data.
[0093] In some embodiments, the processing module 320 is further configured to: perform numerical mapping transformation on the text features in the product feature data and competitor feature data to obtain numerical feature values; and based on preset positive and negative indicators, normalize the numerical feature values corresponding to the positive indicators and normalize the numerical feature values corresponding to the negative indicators after inversion to obtain normalized feature values.
[0094] In some embodiments, the calculation module 330 is further configured to: calculate the information entropy value corresponding to each feature using the entropy weight method based on the numerical feature value and the normalized feature value, thereby obtaining the feature information entropy; calculate the feature weight based on the feature information entropy; and calculate the score of the product and the score of the competitor's product by weighted summation according to the feature weight and the corresponding normalized feature value. In some embodiments, the correction module 350 is further configured to: calculate the mean ratio between the feature values of the product project and the feature values of the competitor project based on the key features selected from the feature values of the product project and the feature values of the competitor project, and use the mean ratio as the feature advantage ratio; extrapolate and correct the integrated value data of the product project based on the feature advantage ratio and the preset correction coefficient to obtain the corrected value of the product project; and constrain the corrected value of the product project to obtain the target value prediction result of the product project.
[0095] In some embodiments, the correction module 350 is further configured to: calculate the upper limit and lower limit of the constraint value of the product project based on the maximum value, minimum value and sample size of the competitor project; constrain the corrected value of the product project between the upper limit and lower limit of the constraint value to obtain the target value prediction result.
[0096] It is understandable that for a detailed description of the real estate project value prediction system 300, please refer to the description of the real estate project value prediction method above.
[0097] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method in any of the above embodiments.
[0098] This application provides a computer program product that includes instructions that, when executed by a processor of a computer device, enable the computer device to perform the steps of the method described in any of the above embodiments.
[0099] Figure 4 A block diagram of an electronic device provided in an embodiment of this application.
[0100] This application provides an electronic device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method in any of the above embodiments.
[0101] like Figure 4 As shown, for ease of understanding, embodiments of this application illustrate a specific electronic device 400.
[0102] Electronic device 400 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 400 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.
[0103] like Figure 4 As shown, device 400 includes a computing unit 401, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 402 or a computer program loaded from storage unit 408 into random access memory (RAM) 403. RAM 403 may also store various programs and data required for the operation of electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. Input / output (I / O) interface 405 is also connected to bus 404.
[0104] Multiple components in electronic device 400 are connected to I / O interface 405. These components include: input unit 406, such as a keyboard or mouse; output unit 407, such as various types of displays or speakers; storage unit 408, such as a disk or optical disk; and communication unit 409, such as a network interface card (NIC), modem, or wireless transceiver. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0105] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods described above. For example, in some embodiments, any one or more of the methods described above can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of any one or more of the methods described above can be performed. Alternatively, in other embodiments, the computing unit 401 can be configured to perform any one or more of the methods described above by any other suitable means (e.g., by means of firmware).
[0106] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this application, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.
[0107] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.
[0108] In the description of this application, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this application, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0109] In the description of this application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.
[0110] Furthermore, the terms "first," "second," etc., used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance, or implicitly specifying the number of technical features indicated in this embodiment. Therefore, features defined with terms such as "first" and "second" in the embodiments of this application can explicitly or implicitly indicate that the embodiment includes at least one of those features. In the description of this application, the word "multiple" means at least two or more, such as two, three, four, etc., unless otherwise explicitly and specifically defined in the embodiments.
[0111] In this application, unless otherwise explicitly specified or limited in the embodiments, the terms "installation," "connection," "joining," and "fixing" appearing in the embodiments should be interpreted broadly. For example, a connection can be a fixed connection, a detachable connection, or an integral part; it can also be a mechanical connection, an electrical connection, etc. Of course, it can also be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication between two components, or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific implementation.
[0112] In this application, unless otherwise expressly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "on top of," and "over" the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.
Claims
1. A method for predicting the value of a real estate project, characterized in that, The method includes: Obtain the product's feature data and competitors' feature data; The product feature data and the competitor product feature data are preprocessed to obtain numerical feature values and normalized feature values; Based on the numerical feature values and the normalized feature values, the entropy weight method is used to calculate the scores of this product and the competitor's product. Based on the product's project score and the competitor's project score, the product's integrated value data is predicted. The integrated value data of the product project is corrected based on the numerical feature values to obtain the target value prediction result of the product project.
2. The method according to claim 1, characterized in that, The method for predicting the integrated value data of the product project based on the product project score and the competitor project score includes: Based on the scores of the product project and the competitor project, the first predicted value data of the product project is calculated by Gaussian kernel function, and the second predicted value data of the product project is calculated by linear extrapolation. The fusion weight is determined based on the score range relationship between the product's score and the competitor's score. The first predicted value data and the second predicted value data are weighted and fused based on the fusion weight to obtain the fused value data of the product project.
3. The method according to claim 2, characterized in that, The process of calculating the first predicted value data of the product project based on the product project score and the competitor product project score using a Gaussian kernel function, and calculating the second predicted value data of the product project using linear extrapolation, includes: Calculate the absolute distance between the scores of this product and the competitor's product based on the product's score and the competitor's score. The absolute distance of the score is normalized to obtain the normalized distance; Based on the normalized distance and the Gaussian kernel function, the value weights of competing projects are calculated. The first predicted value data is obtained by weighting and summing the values of the competing projects based on their respective value weights. The competitor projects are sorted according to their scores, and the score of the product project is compared with the scores of the sorted competitor projects to determine the second predictive value data.
4. The method according to claim 3, characterized in that, The process of sorting competitor projects according to their scores and comparing the score of the product project with the scores of the sorted competitor projects to determine the second predicted value data includes: The competitor projects are sorted in ascending order according to their scores to obtain the competitor project score range, wherein the competitor project score range includes the maximum value and the minimum value of the range; The sensitivity slope is calculated based on the scores and values of the competing products within the score range of the competing products, where the score relationship is a preset positional relationship. Determine the position of the product's score within the score range of the competitor's product, and determine the second predicted value data based on different positional relationships.
5. The method according to claim 4, characterized in that, The step of determining the position of the product's score within the score range of competing products, and determining the second predicted value data based on different positional relationships, includes: If the score of the product item is higher than the maximum value of the interval, then based on the sensitivity slope, the score of the competitor item corresponding to the maximum value of the interval, and the value of the competitor item, a linear extrapolation is performed to the first side to obtain the second predicted value data, wherein the first side represents the side with a score higher than the maximum value of the interval; If the score of the product item is lower than the minimum value of the interval, then based on the sensitivity slope, the score of the competitor item corresponding to the minimum value of the interval, and the value of the competitor item, a linear extrapolation is performed to the second side to obtain the second predicted value data, wherein the second side represents the side with a score lower than the minimum value of the interval; If the score of the product item falls within the score range of the competitor item, then based on the sorted scores and values of the competitor items, the score of the product item is linearly interpolated to obtain the second predicted value data.
6. The method according to claim 1, characterized in that, The preprocessing of the product feature data and competitor feature data to obtain numerical feature values and normalized feature values includes: The text features in the product feature data and the competitor product feature data are numerically mapped and transformed to obtain the numerical feature values. Based on preset positive and negative indicators, the numerical feature values corresponding to the positive indicators are normalized, and the numerical feature values corresponding to the negative indicators are inverted and then normalized to obtain the normalized feature values.
7. The method according to claim 1, characterized in that, The calculation of the product score and competitor score using the entropy weight method based on the numerical feature values and the normalized feature values includes: Based on the numerical feature values and the normalized feature values, the information entropy value corresponding to each feature is calculated using the entropy weight method to obtain the feature information entropy; Based on the feature information entropy, the feature weights are calculated; Based on the feature weights and the corresponding normalized feature values, the scores of the product item and the competitor item are calculated by weighted summation.
8. The method according to claim 1, characterized in that, The numerical feature values include feature values of this product project and feature values of competing product projects; the step of correcting the integrated value data of this product project based on the numerical feature values to obtain the target value prediction result of this product project includes: Based on the key features selected from the feature values of the product and the feature values of the competitors, the mean ratio between the feature values of the product and the feature values of the competitors is calculated, and the mean ratio is used as the feature advantage ratio. Based on the aforementioned feature advantage ratio and preset correction coefficient, the integrated value data of the product project is extrapolated and corrected to obtain the corrected value of the product project. The modified value of the product project is constrained to obtain the target value prediction result of the product project.
9. The method according to claim 8, characterized in that, The constraint processing on the modified value of the product item to obtain the target value prediction result of the product item includes: Based on the maximum and minimum values of competing projects and the sample size, calculate the upper and lower bounds of the constraint value of this project. The target value prediction result is obtained by constraining the modified value of the product item between the upper limit and the lower limit of the constraint.
10. A real estate project value prediction system, characterized in that, The system includes: The acquisition module is used to acquire the product's feature data and competitors' feature data; The processing module is used to preprocess the product feature data and the competitor product feature data to obtain numerical feature values and normalized feature values; The calculation module is used to calculate the score of this product and the score of the competitor's product using the entropy weight method based on the numerical feature value and the normalized feature value. The prediction module is used to predict the integrated value data of the product project based on the product project score and the competitor project score. The correction module is used to correct the integrated value data of the product project based on the numerical feature values to obtain the target value prediction result of the product project.