A housing demand dynamic prediction method based on three-drive linkage

Abstract

The application discloses a kind of based on three drive linkage's housing demand dynamic prediction method, including multi-source data acquisition and index extraction, construct basic demand generation submodel, construct payment capacity conversion submodel, construct expectation and policy adjustment submodel, three drive linkage simultaneous and iterative solution, output prediction result.The application constructs basic demand generation module, payment capacity conversion module and expectation policy adjustment module, and the three modules are simultaneously solved into system dynamics equation set by endogenous house price feedback loop and expectation acceleration loop, iterative solution is obtained using discrete time step, the housing demand of each prediction period is obtained, high-precision dynamic prediction is realized, by endogenization house price and mood, so that model can independently calculate the chain reaction of demand boom, recession, policy transmission, can also accurately inject policy shock variable, observe its through two paths of improving payment capacity, boost mood complete transmission process of amplifying demand, provide quantitative tool for policy effectiveness evaluation.

Description

Technical Field

[0001] This invention relates to the field of data analysis and market forecasting technology, specifically to a dynamic quantitative forecasting method and system for housing demand that integrates population structure changes, economic affordability, and policy expectation adjustments. Background Technology

[0002] Housing demand forecasting is a core basis for urban planning and real estate market regulation. Traditional forecasting methods mainly rely on the following three types:

[0003] First, the population trend extrapolation method uses indicators such as urbanization rate and the number of new households to estimate the total future housing demand through linear fitting. This method assumes that demand is unidirectionally and fixedly correlated with population, ignoring the filtering and disturbance of housing behavior by economic costs and market sentiment.

[0004] Secondly, the econometric model method uses multiple linear regression or vector autoregression to fit historical transaction volumes with income, interest rates, housing prices, etc. as independent variables. These models are mostly static structures and cannot express the dynamic feedback loop between variables, especially the closed loop of price changes affecting demand itself.

[0005] Third, the policy effect evaluation method usually uses dummy variables to represent policy shocks; this method assumes that the effectiveness is fixed and does not consider the multiplier effect and time-varying decay of the policy by changing expectations.

[0006] The common flaw in the aforementioned methods lies in treating the three major demand drivers—population, economy, and policy—as isolated or additive relationships, failing to construct an organic predictive system where these three factors interact and provide endogenous feedback. Specifically: they cannot simulate the "demand crowding-out" effect—where population influx triggers rising housing prices, thereby weakening the purchasing power of new urban residents; they cannot capture the accelerating mechanism of "demand overdraft"—where favorable policies improve market sentiment, leading to the premature release of pent-up demand; and they cannot reveal the "demand freeze" phenomenon—where weakening housing price expectations indefinitely postpone demand for improved housing. Therefore, there is an urgent need for a dynamic housing demand prediction scheme that can systematically describe the aforementioned interconnected mechanisms and be implemented based on a computable model. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a dynamic housing demand forecasting method based on the linkage of three drivers, aiming to solve the technical problem that existing forecasting methods cannot characterize the dynamic linkage and endogenous feedback among the three drivers of population, economy, and policy.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] A method for dynamic forecasting of housing demand based on three-drive linkage includes the following steps:

[0010] S1. Multi-source data collection and indicator extraction: Obtain population-driven indicators, economic-driven indicators, and expectation and policy-driven indicators for the target region;

[0011] S2. Construct a basic demand generation sub-model: Generate basic demand sequences for each forecast period based on population-driven indicators;

[0012] S3. Construct a payment capacity conversion sub-model: Using the house price-to-income ratio and credit availability index as inputs, calculate the conversion rate of basic demand into effective demand through a nonlinear function. The house price parameter in the house price-to-income ratio is endogenously determined by the supply-demand ratio of the previous period.

[0013] S4. Construct a sub-model for expectations and policy adjustment: Generate a market sentiment index based on expectations and policy-driven indicators, and map the market sentiment index to the acceleration rate of demand release;

[0014] S5. Three-Drive Linkage and Iterative Solution: The basic demand generation sub-model, the payment ability transformation sub-model, and the expectation and policy adjustment sub-model are linked through the endogenous housing price feedback equation and the sentiment transmission equation to form a three-drive linkage dynamic equation system.

[0015] Set initial boundary conditions, iterate and solve for each prediction time step until the endogenous house price converges, and obtain the final housing demand forecast and endogenous house price for that time step.

[0016] S6. Output the prediction results.

[0017] Preferably, the payment capability transformation sub-model adopts an S-shaped logic function:

[0018] Con(t)=1 / [1+exp(γ×(HPI(t)-HPI_crit) / Cr_avail(t))],

[0019] Wherein, Con(t) is the conversion rate, HPI(t) is the price-to-income ratio, HPI_crit is the critical price-to-income ratio, γ is the steepness parameter, and Cr_avail(t) is the credit availability index. The house price in HPI(t) is the predicted house price P_pred derived endogenously from the previous iteration or the previous time step.

[0020] Preferably, the market sentiment index Sent(t) in the expectation and policy adjustment sub-model is calculated in the following way:

[0021] Sent(t) = α × Sent(t-1) + (1-α) × (k1 × search sentiment change term + k2 × policy shock term - k3 × credit risk adjustment term),

[0022] Where α is the smoothing coefficient, the search sentiment change item is represented by the year-on-year change rate of the target keyword search index, and the policy impact item is a weighted composite of the change in the intensity of the purchase and sale restriction policy and the subsidy intensity score.

[0023] Preferably, the simultaneous and iterative solution further includes: updating the housing price based on the previous period's housing price and the current supply-demand ratio, feeding the updated housing price back to the affordability conversion sub-model, and feeding the emotional changes caused by housing price changes back to the expectation and policy adjustment sub-model, and repeating the process until the housing price change is less than a preset threshold.

[0024] Preferably, the prediction results output also includes: setting multiple different scenarios of population inflow, income growth and policy strength, running the iterative solution for each scenario, and outputting the housing demand prediction range under each scenario and the structure decomposed by first-time homebuyer demand and improvement demand.

[0025] Preferably, a dynamic housing demand forecasting system based on three-drive linkage includes:

[0026] The data acquisition unit is used to acquire population-driven indicators, economic-driven indicators, and expectation and policy-driven indicators for the target area.

[0027] The model storage and solution unit stores the simultaneous model and iterative solution program constructed by the above method, and calls the data to perform prediction calculations;

[0028] The visualization output unit is used to graphically display the predicted demand, endogenous housing prices, and the contribution of each driving force.

[0029] Preferably, a computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) This invention constructs a basic demand generation module, a payment capacity conversion module and an expected policy adjustment module, and combines the three modules into a system dynamic equation set through an endogenous housing price feedback loop and an expected acceleration loop. The system dynamic equation set is solved iteratively using discrete time steps to obtain the housing demand for each prediction period, thereby achieving high-precision dynamic prediction.

[0032] (2) By endogenizing housing prices and sentiment, this invention enables the model to autonomously calculate the chain reaction of demand boom, recession and policy transmission, thus eliminating the rigidity of static forecasting.

[0033] (3) This invention can accurately inject policy shock variables such as “lower down payment”, “lower interest rate”, and “housing subsidy”, and observe the complete transmission process of demand amplification through two paths: improving payment ability and boosting sentiment, providing a quantitative tool for policy effectiveness assessment;

[0034] (4) This invention can issue a structural divergence warning of “population rising and payment capacity falling” by real-time decomposition of the contribution of driving forces, indicating the risk of demand gap. Attached Figure Description

[0035] Figure 1 This is an overall flowchart of the method of the present invention. Detailed Implementation

[0036] 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.

[0037] like Figure 1 As shown, the present invention provides a technical solution: a method for dynamic prediction of housing demand based on three-drive linkage, comprising the following steps:

[0038] S1. Multi-source data collection and indicator extraction: Acquire historical and current data of the target area, and extract population-driven indicator groups, economic-driven indicator groups, and expectation and policy-driven indicator groups;

[0039] S2. Construct a basic demand generation sub-model: Based on the queue element method, according to population-driven indicators, the number of potential housing-demanding families in each period is generated based on age-based population and family separation tendency, as a basic demand sequence.

[0040] S3. Construct a payment capacity conversion sub-model: Using the house price-to-income ratio and credit availability index as inputs, calculate the conversion rate of basic demand into effective demand through a nonlinear logic function. The house price parameter in the house price-to-income ratio is endogenously determined by the supply-demand ratio of the previous period.

[0041] S4. Construct a sub-model for expectations and policy adjustment: Generate a market sentiment index based on expectations and policy-driven indicators, and map the market sentiment index to the acceleration rate of demand release;

[0042] S5. Three-Drive Linkage and Iterative Solution: The basic demand generation sub-model, the payment ability transformation sub-model, and the expectation and policy adjustment sub-model are linked through the endogenous housing price feedback equation and the sentiment transmission equation to form a three-drive linkage dynamic equation system.

[0043] Set initial boundary conditions, iterate and solve for each prediction time step until the endogenous house price converges, and obtain the final housing demand forecast and endogenous house price for that time step.

[0044] S6. Output forecast results: Set different combinations of population, income and policy strength assumptions, repeat step S5, and output the demand forecast interval and structural decomposition under multiple scenarios;

[0045] The payment capability transformation sub-model adopts an S-shaped logic function:

[0046] Con(t)=1 / [1+exp(γ×(HPI(t)-HPI_crit) / Cr_avail(t))],

[0047] Where Con(t) is the conversion rate, HPI(t) is the price-to-income ratio, HPI_crit is the critical price-to-income ratio, γ is the steepness parameter, Cr_avail(t) is the credit availability index, and the house price in HPI(t) is the predicted house price P_pred derived endogenously from the previous iteration or the previous time step.

[0048] The market sentiment index Sent(t) in the expectation and policy adjustment sub-model is calculated as follows:

[0049] Sent(t) = α × Sent(t-1) + (1-α) × (k1 × search sentiment change term + k2 × policy shock term - k3 × credit risk adjustment term),

[0050] Where α is the smoothing coefficient, the search sentiment change item is represented by the year-on-year change rate of the target keyword search index, and the policy impact item is a weighted composite of the change in the intensity of the purchase and sale restriction policy and the subsidy intensity score.

[0051] The simultaneous and iterative solution further includes: updating the housing price based on the previous period's housing price and the current supply-demand ratio, feeding the updated housing price back to the affordability conversion sub-model, and feeding the sentiment changes caused by housing price changes back to the expectation and policy adjustment sub-model, and repeating the process until the housing price change is less than a preset threshold.

[0052] The forecast output also includes: setting multiple different scenarios of population inflow, income growth, and policy strength, running iterative solutions for each scenario, and outputting the housing demand forecast range under each scenario and the structure of decomposition by first-time homebuyer demand and improvement demand;

[0053] A dynamic housing demand forecasting system based on three-drive linkage includes:

[0054] The data acquisition unit is used to acquire population-driven indicators, economic-driven indicators, and expectation and policy-driven indicators for the target area.

[0055] The model storage and solution unit stores the simultaneous model and iterative solution program constructed by the above methods, and calls the data to perform prediction calculations;

[0056] The visualization output unit is used to graphically display the predicted demand, endogenous housing prices, and the contribution of each driving force.

[0057] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the above-described method.

[0058] Example

[0059] This embodiment provides a dynamic housing demand forecasting method based on three-drive linkage, which forecasts housing demand for the next 12 quarters using quarters as the time step, including the following steps:

[0060] Step S1: Multi-source data collection and indicator extraction:

[0061] The following data was obtained from government statistical yearbooks, the online contract signing and filing system of the housing and construction department, mainstream real estate information platforms, and internet search engines through API interfaces and customized web crawlers:

[0062] Population-driven indicators: resident population by age group P_age, urbanization rate Urb, average household size H_size, number of marriage registrations per quarter Mar, number of divorce registrations per quarter Div;

[0063] Economic driving indicators: per capita disposable income of urban residents Inc, weighted average interest rate of housing loans r, approved pre-sale area of ​​newly built commercial housing S_new, and average listing price of existing housing P_list;

[0064] Expectations and policy indicators: Search index SI for the keyword "buying a house", Policy_s (a standardized value between 0 and 1 obtained through text analysis, where 0 represents no restriction and 1 represents complete prohibition), Sub_s (standardized to 0-1), and NPL (non-performing loan ratio for personal housing loans).

[0065] Step S2: Construct a basic requirement generation sub-model;

[0066] The queue element method is used to calculate the base of newly added household housing demand D_base(t) in the current period;

[0067] First, calculate the net number of households formed:

[0068] ΔHH(t)=

[0069] [P20-34(t)×α_mar(t)+Div(t)×β_split] / H_size_adjust-ΔHH_vacancy(t),

[0070] Among them, P20-34(t) represents the number of the main marriage and childbearing population aged 20-34;

[0071] α_mar(t) is the marriage rate for this age group, which is derived from the moving average of Mar(t) / P20-34(t);

[0072] β_split is the coefficient for the division of property due to divorce, which can be taken as 1.2;

[0073] H_size_adjust is used to adjust the household size to take into account the trend of living alone;

[0074] ΔHH_vacancy(t) represents the number of housing units vacated due to death or population outflow;

[0075] Basic requirements are defined as follows:

[0076] D_base(t)=ΔHH(t)×θ_new+HH_stock(t)×δ_improve,

[0077] Where θ_new represents the home-buying tendency of new families (first-time homebuyers), and δ_improve represents the replacement rate of existing families (estimated based on changes in family size and building age).

[0078] Step S3: Construct a payment capability transformation sub-model;

[0079] This step transforms D_base(t) into valid demand D_eff(t) with payment capacity;

[0080] First, define the house price-to-income ratio index:

[0081] HPI(t)=P_pred(t)×Area_std / (Inc(t)×HH_size(t)),

[0082] Where P_pred(t) is the model-inherent predicted house price (see step S5), and Area_std is the standard housing area (taken as 90 square meters).

[0083] The initial value of P_pred(t) is taken from the current period's P_list;

[0084] The credit availability index Cr_avail(t) is a decreasing function of the interest rate r(t) and the lower limit of the down payment ratio, which can be simplified to: Cr_avail(t) = 2 - e^(0.05 × r(t)), normalized to [0.1, 1].

[0085] The payment capacity conversion rate Con(t) is calculated using a sigmoid logic function:

[0086] Con(t)=1 / [1+exp(γ×(HPI(t)-HPI_crit) / Cr_avail(t))],

[0087] Where HPI_crit is the critical house price-to-income ratio (e.g., 6), and γ is the steepness parameter (2).

[0088] When HPI is far above the critical level and credit is tight, Con(t) approaches 0, and a large amount of basic demand cannot be converted.

[0089] Conversely, it tends to 1;

[0090] The effective demand is: D_eff(t) = D_base(t) × Con(t);

[0091] Step S4: Construct the sub-model of expectations and policy adjustment:

[0092] Generate the market sentiment index Sent(t) and the demand release acceleration factor Acc(t);

[0093] First, calculate the search sentiment term: Se(t) = [SI(t) - SI(t-4)] / SI(t-4) (year-on-year change rate);

[0094] Policy shock term: Policy_shock(t) = w1 × [Policy_s(t) - Policy_s(t-1)] + w2 × Sub_s(t), where w1 and w2 are weights, reflecting the different impacts of relaxing purchase restrictions and direct subsidies;

[0095] Historically accumulated sentiment is expressed using an exponentially weighted moving average:

[0096] Sent(t) = 0.6 × Sent(t-1) + 0.4 × (k1 × Se(t) + k2 × Policy_shock(t) - k3 × ΔNPL(t)), where k1, k2, and k3 are influence coefficients;

[0097] The acceleration factor Acc(t) is linearly mapped to Sent(t) and is limited:

[0098] Acc(t) = clamp(1 + ρ × Sent(t), 0.5, 1.5), where ρ is the scaling factor;

[0099] When emotions are extremely optimistic, the acceleration factor can reach 1.5, which represents an overdraft of demand;

[0100] In a pessimistic scenario, the value drops to 0.5, indicating frozen demand.

[0101] Step S5: Three-drive linkage equation and iterative solution:

[0102] The core of this invention is to solve a system of equations and introduce a closed loop.

[0103] For the prediction period from t=1 to T, perform the following iterations:

[0104] 1. Set the initial house price guess for period t as P_pred_guess = P_list(t-1) (the actual average market price lagged by one period);

[0105] 2. Calculate HPI using P_pred_guess, then calculate Con(t), and finally obtain D_eff(t);

[0106] 3. Calculate the final demand after accelerated release: D_final(t) = D_eff(t) × Acc(t);

[0107] 4. Calculate the market supply-demand ratio: Supply-demand ratio = D_final(t) / S_new(t);

[0108] 5. Update endogenous housing prices using the housing price adjustment equation: P_pred_new=P_list(t-1)×[1+η×(supply-demand ratio-1)], where η is the price adjustment coefficient (e.g., 0.2), indicating that housing prices rise when supply is less than demand;

[0109] 6. Determine if |P_pred_new - P_pred_guess| < ε (e.g., 0.1%). If convergence occurs, record P_pred(t) = P_pred_new and D(t) = D_final(t).

[0110] Otherwise, set P_pred_guess = P_pred_new and jump back to step 2;

[0111] Meanwhile, the Sent(t) function in the expectation module will also use the previous period's housing price change ΔP(t-1) and historical sentiment to achieve cross-period linkage;

[0112] All time steps are solved sequentially, starting from the initial known data and progressing quarter by quarter.

[0113] Step S6: Multi-scenario prediction output:

[0114] Three scenarios—baseline, optimistic, and pessimistic—are set, with differentiated settings for future population inflow growth, income growth, and the extent of relaxation of purchase restrictions.

[0115] Run the solver on the model to obtain the demand sequence under each scenario and plot the envelope curve;

[0116] Meanwhile, the model can be decomposed to show the contributions of D_base(t), Con(t), and Acc(t) at each period, revealing the driving structure;

[0117] The prediction system corresponding to the above method includes:

[0118] The data acquisition unit collects multi-source data periodically through web crawlers and API interface modules and stores it in a time-series database;

[0119] The model storage and solution unit is a computing server equipped with a processor and memory. The memory stores the model equations and iterative solution engine program from S2 to S5 above. The processor executes the program, calls up the data, and outputs the prediction results.

[0120] The visualization output unit generates interactive charts from the prediction results, displaying the contribution of each driving force, demand forecast curves, and early warning signals.

[0121] It should be noted that, in this document, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0122] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for dynamic prediction of housing demand based on three-drive linkage, characterized in that: Includes the following steps: S1. Multi-source data collection and indicator extraction: Obtain population-driven indicators, economic-driven indicators, and expectation and policy-driven indicators for the target region; S2. Construct a basic demand generation sub-model: Generate basic demand sequences for each forecast period based on population-driven indicators; S3. Construct a payment capacity conversion sub-model: Using the house price-to-income ratio and credit availability index as inputs, calculate the conversion rate of basic demand into effective demand through a nonlinear function. The house price parameter in the house price-to-income ratio is endogenously determined by the supply-demand ratio of the previous period. S4. Construct a sub-model for expectations and policy adjustment: Generate a market sentiment index based on expectations and policy-driven indicators, and map the market sentiment index to the acceleration rate of demand release; S5. Three-Drive Linkage and Iterative Solution: The basic demand generation sub-model, the payment ability transformation sub-model, and the expectation and policy adjustment sub-model are linked through the endogenous housing price feedback equation and the sentiment transmission equation to form a three-drive linkage dynamic equation system. Set initial boundary conditions, iterate and solve for each prediction time step until the endogenous house price converges, and obtain the final housing demand forecast and endogenous house price for that time step. S6. Output the prediction results.

2. The method for dynamic prediction of housing demand based on three-drive linkage according to claim 1, characterized in that: The payment capability transformation sub-model adopts an S-shaped logic function: Con(t)=1 / [1+exp(γ×(HPI(t)-HPI_crit) / Cr_avail(t))], Wherein, Con(t) is the conversion rate, HPI(t) is the price-to-income ratio, HPI_crit is the critical price-to-income ratio, γ is the steepness parameter, and Cr_avail(t) is the credit availability index. The house price in HPI(t) is the predicted house price P_pred derived endogenously from the previous iteration or the previous time step.

3. The method for dynamic prediction of housing demand based on three-drive linkage according to claim 1, characterized in that: The market sentiment index Sent(t) in the expectation and policy adjustment sub-model is calculated as follows: Sent(t) = α × Sent(t-1) + (1-α) × (k1 × search sentiment change term + k2 × policy shock term - k3 × credit risk adjustment term), Where α is the smoothing coefficient, the search sentiment change item is represented by the year-on-year change rate of the target keyword search index, and the policy impact item is a weighted composite of the change in the intensity of the purchase and sale restriction policy and the subsidy intensity score.

4. The method for dynamic prediction of housing demand based on three-drive linkage according to claim 1, characterized in that: The simultaneous and iterative solution further includes: updating the housing price based on the previous period's housing price and the current supply-demand ratio, feeding the updated housing price back to the affordability conversion sub-model, and feeding the sentiment changes caused by housing price changes back to the expectation and policy adjustment sub-model, and repeating this process until the housing price change is less than a preset threshold.

5. The method for dynamic prediction of housing demand based on three-drive linkage according to claim 1, characterized in that: The prediction results output also includes: setting multiple different scenarios of population inflow, income growth, and policy strength, running the iterative solution for each scenario, and outputting the housing demand prediction range under each scenario and the structure decomposed by first-time homebuyer demand and improvement demand.

6. A dynamic housing demand forecasting system based on three-drive linkage, characterized in that, include: The data acquisition unit is used to acquire population-driven indicators, economic-driven indicators, and expectation and policy-driven indicators for the target area. The model storage and solution unit stores the simultaneous model and iterative solution program constructed by the method described in any one of claims 1 to 5, and calls the data to perform prediction calculations; The visualization output unit is used to graphically display the predicted demand, endogenous housing prices, and the contribution of each driving force.

7. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the method as described in any one of claims 1 to 5.

Images