A car sales price prediction method based on multi-source data fusion

Abstract

The application relates to the technical field of automobile sales data processing, and discloses an automobile sales price prediction method based on multi-source data fusion. The method establishes an automobile sales price prediction session, registers vehicle objects, vehicle configurations, sales regions, sales channels, observation time and version information; according to source records, price fields, expense items, business document states and record occurrence time, price source labels, price caliber labels, sales stage labels and stage comparable price groups are formed; according to manufacturer price adjustment, promotion change, regional subsidy, inventory pressure, competitive product price reduction, financial policy change and vehicle condition disclosure records, price events and influence windows are registered, and event influence labels are written; the stage comparable price groups are subjected to source deduplication, expense composition, transaction closing, vehicle condition credibility, inventory state and regional channel verification, and effective sample areas, to-be-verified areas and isolation areas are written; and automobile sales price prediction versions are generated according to the sample areas and states.

Description

Technical Field

[0001] This invention relates to the field of automobile sales data processing technology, specifically to a method for predicting automobile sales prices based on multi-source data fusion. Background Technology

[0002] Forecasting car sales prices typically requires simultaneously reading basic vehicle information, historical transaction records, platform listing records, dealer price records, inventory records, promotional policy records, financing plan records, trade-in subsidy records, and regional transaction records. While all of these data are related to vehicle prices, their calculation methods are inconsistent: platform listing prices often reflect the willingness to display the vehicle, actual transaction prices reflect the final confirmation from both the buyer and seller, financing installment prices may include interest, service fees, or subsidy discounts, trade-in prices may include old vehicle subsidies, and inventory clearance prices are often affected by inventory age and regional inventory pressure. Directly using these price data as a single price sample for forecasting can easily lead to a lack of comparability between the underlying forecast data.

[0003] Existing car price prediction methods typically use fields such as vehicle model, age, mileage, configuration, region, historical transaction price, page views, and inquiry volume as input, and output predicted prices through statistical rules, valuation models, or machine learning models. These methods usually focus on the number of features, model structure, or prediction error, but rarely conduct unified verification of the price composition, sales stage, validity period, channel source, and credibility of prices from different sources before prediction. When the same model generates multiple price records on different platforms, at different dealerships, and during different promotional periods, the system struggles to distinguish which prices are valid transaction prices that can be directly used for prediction, and which prices are merely listing prices, promotional prices, inventory clearance prices, or abnormal quotes.

[0004] Especially after events such as new car price reductions, regional inventory changes, manufacturer subsidy switching, dealer promotion adjustments, and supplementary disclosure of vehicle condition information, vehicle sales prices can fluctuate significantly in a short period. Existing methods that directly update prediction results based on historical samples are prone to mixing outdated high-price samples, short-term promotional low-price samples, financially converted prices, and quotes from different channels, leading to predicted prices deviating from the current sales scenario. Therefore, in scenarios involving cross-channel car sales, inventory vehicle pricing, and regional price linkage, a car sales price prediction method is still needed that can verify the accuracy of multi-source price records, identify event validity periods, isolate abnormal prices, and handle sales stage attribution before price prediction. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for predicting automobile sales prices based on multi-source data fusion, thus solving the aforementioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting automobile sales prices based on multi-source data fusion, comprising:

[0007] S1. Establish a car sales price forecasting session, register vehicle objects, model configurations, sales regions, sales channels, observation time, data cutoff time, price caliber version, sales stage version, and forecast output version, and read vehicle basics, inventory, listings, quotations, orders, transactions, financing, trade-ins, promotions, vehicle condition, and regional transaction records.

[0008] S2. Based on the source record, price field, expense item, business document status, sales channel and record occurrence time, form price source label, price caliber label and sales stage label, and form stage comparable price groups according to vehicle object, price caliber, sales stage, regional channel and price validity period;

[0009] S3. Based on manufacturer price adjustments, promotional changes, regional subsidies, inventory pressure, competitor price reductions, changes in financial policies, and vehicle condition disclosure records, register price events and their impact windows, and write event impact markers to the corresponding price records.

[0010] S4. Perform source deduplication, cost composition verification, transaction closure verification, vehicle condition verification, inventory status verification, and regional channel verification on the comparable price group of the stage, and write the verification status into the valid sample area, the pending verification area, and the isolation area.

[0011] S5. Based on the valid sample area, event impact markers, vehicle condition reliability, inventory status, and regional channel status, generate a version of the automobile sales price forecast that corresponds to the forecast purpose, price caliber, sales stage, and future price observation period.

[0012] Furthermore, in S1:

[0013] Vehicle objects are formed by registering individual vehicle objects, vehicle model objects, and regional vehicle model objects. Individual vehicle objects are jointly defined by vehicle identification number, internal inventory number, batch number upon entry, and store inventory number. Vehicle model objects are jointly defined by brand, series, model, year, configuration name, energy type, and drive type. Regional vehicle model objects are based on vehicle model objects and include sales region and sales channel.

[0014] Furthermore, in S1:

[0015] The observation period consists of historical observation time and future price observation period. The historical observation time is read backward from the data cutoff time. The future price observation period is registered according to the predicted purpose, which includes the corresponding purpose in inventory quotation, customer inquiry, order price locking, clearance disposal, wholesale disposal and cross-regional transfer.

[0016] Furthermore, in S2:

[0017] The source record is formed by the data source, receiving batch, vehicle object, record occurrence time, record receiving time and original serial number. When the original serial number is missing, a temporary source record is formed. The temporary source record enters the verification area, and after the original serial number is supplemented, the process of forming price source label, price caliber label and sales stage label continues according to the supplemented source record.

[0018] Furthermore, in S2:

[0019] Price source tags are formed based on business document type, source platform, price field name, order number, quotation number, financial plan number, trade-in record number, payment status, and delivery status. Actual transaction source is jointly registered by transaction order number, payment status, and delivery status. Dealer quotation source is jointly registered by quotation number, sales consultant quotation record, and customer lead number. Platform listing source is jointly registered by vehicle source release number, display price, listing time, and platform number.

[0020] Furthermore, in S2:

[0021] The price caliber label is formed based on the price field and the expense items. The expense items include purchase tax, insurance premium, license plate fee, decoration fee, financial service fee, loan interest, extended warranty fee, delivery fee, logistics fee, preparation fee, trade-in subsidy, manufacturer cash discount, regional subsidy and dealer discount. When an expense item is missing, the corresponding price record is recorded as expense item pending.

[0022] Furthermore, in S2:

[0023] Sales stage tags are formed based on the status of business documents, the time of record occurrence, order status, inventory status, customer lead status, and sales channel. The sales stage tags are one of the following: vehicle source filing stage, platform display stage, customer inquiry stage, in-store negotiation stage, order price locking stage, delivery and transaction stage, order cancellation and rollback stage, wholesale disposal stage, and clearance disposal stage.

[0024] When both the sales stage field and the business document status are missing, the corresponding price record is registered as a stage pending verification status.

[0025] Furthermore, in S3:

[0026] The impact window of a price event is formed according to the event type. The impact window of a manufacturer's price adjustment event starts from the effective time of the price adjustment and ends before the effective time of the next manufacturer's suggested retail price version. The impact window of a promotional change event starts from the effective time of the promotion and ends at the expiration time of the promotion. The impact window of an inventory pressure event starts when the number of inventory units reaches the inventory pressure control value or when the inventory vehicles enter a high inventory age state and ends when the inventory pressure relief value continuously meets two effective sales cycles.

[0027] Furthermore, in S4:

[0028] The valid sample area accepts price records that are complete in terms of price scope, clear in terms of sales stage, have not expired in terms of price validity period, complete in terms of transaction loop, reliable in terms of vehicle condition, have available inventory and are consistent in terms of regional channels.

[0029] Price records for the pending verification area, including items for receiving fees pending verification, business types pending verification, stages pending verification, invoices pending verification, vehicle conditions pending verification, warehousing time pending verification, regional channels pending verification, and subsidy eligibility pending verification;

[0030] The quarantine zone receives price records for price conflicts, transaction price conflicts, pending verification of objects, order cancellations and rollbacks, inventory anomalies, samples before vehicle condition disclosure, and invalid sources.

[0031] Furthermore, in S5:

[0032] The automobile sales price forecast version registration includes the forecast output version, sample version number, price caliber, sales stage, forecast purpose, future price observation period, event impact marker, vehicle condition reliability status, inventory status, regional channel status, and version status.

[0033] The version status is one of the following: pending activation, effective, affected by the event, under observation pending verification, isolated from an anomaly, rolled back to effective, or no effective price.

[0034] Compared with existing technologies, this invention provides a method for predicting automobile sales prices based on multi-source data fusion, which has the following beneficial effects:

[0035] 1. This invention addresses the technical problem of incomparable prices being mixed into the forecast sample due to inconsistencies in price calibers from different sources, sales stages, and event impact times in automobile sales price forecasting. The invention establishes an automobile sales price forecasting session in S1, fixing vehicle objects, model configurations, regional channels, observation time, and version boundaries. In S2, it generates price source labels, price caliber labels, sales stage labels, and comparable price groups for each stage. In S3, it registers price events and their impact windows, such as manufacturer price adjustments, promotional changes, regional subsidies, inventory pressure, competitor price reductions, changes in financial policies, and vehicle condition disclosures. Finally, in S4, it processes the price records according to... The verification status is written into the valid sample area, the verification area, and the isolation area. Then, S5 generates a car sales price prediction version corresponding to the prediction purpose, price caliber, sales stage, and future price observation period. This allows for the differentiation of the applicable boundaries of platform listing price, dealer quotation, order lock-in price, actual transaction price, financial plan price, trade-in subsidy price, wholesale transaction price, and inventory clearance price before prediction. This avoids the direct mixing of price records under different calibers, stages, channels, and event influences, and ensures that the car sales price prediction version has clear data sources, price calibers, sales stages, event influences, and sample status basis.

[0036] 2. This invention changes the price forecasting process from simply reading historical prices and outputting predicted values ​​to first performing source deduplication, cost composition verification, transaction closure verification, vehicle condition credibility verification, inventory status verification, and regional channel verification. Then, it generates a forecast version based on the effective sample area, event impact markers, vehicle condition credibility status, inventory status, and regional channel status. For situations that may cause short-term price fluctuations, such as new car price adjustments, promotion switching, changes in regional subsidies, inventory backlog, competitor price reductions, changes in financial policies, and supplementary vehicle condition disclosures, it can limit the affected price records to the corresponding impact window and corresponding sample partition. This reduces the interference of expired high-price samples, short-term low-price samples used for attracting traffic, canceled and rolled-back prices, vehicle condition pending verification prices, and inconsistent regional channel prices on the forecast results. This makes the generated car sales price forecast version more suitable for different sales use scenarios such as inventory quoting, customer inquiries, order price locking, clearance disposal, wholesale disposal, and cross-regional transfer. Attached Figure Description

[0037] Figure 1 This is a schematic diagram of the process of a method for predicting automobile sales prices based on multi-source data fusion according to the present invention. Detailed Implementation

[0038] 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.

[0039] Example: Figure 1 A method for predicting automobile sales prices based on multi-source data fusion is presented, including:

[0040] S1. Establish a car sales price forecasting session, registering vehicle targets, model configurations, sales regions, sales channels, observation periods, data cutoff times, price caliber versions, sales stage versions, and forecast output versions. It also reads vehicle basic data, inventory, listings, quoted prices, orders, transactions, financing, trade-ins, promotions, vehicle condition, and regional transaction records. Specific implementation details are as follows:

[0041] First, establish a car sales price forecasting session. This session defines the scope of vehicles, time frame, price caliber, sales stage, and output version used in this price forecast. When establishing the session, register the session number, vehicle type, model configuration, sales region, sales channel, observation period, data cutoff time, price caliber version, sales stage version, and forecast output version.

[0042] Vehicle objects are used to distinguish between single vehicle objects, vehicle model objects, and regional vehicle model objects. A single vehicle object is registered together with the vehicle identification code, internal inventory number, inbound batch number, and store inventory number. When the vehicle identification code is missing, it will not enter the single vehicle price prediction process, but will only enter the vehicle model object price prediction process, and the vehicle object will be registered as a vehicle identification code pending supplementation state. When the internal inventory number and the vehicle identification code do not correspond, the vehicle object will be registered as an object pending verification state. Price records in the object pending verification state will not enter the subsequent valid sample area.

[0043] Vehicle configuration is registered by brand, series, model, year, configuration name, energy type, drive type, number of seats, and key configuration version. When the configuration name is missing, the vehicle announcement model, sales name, and inventory file name are read for consistency verification. If the three cannot form the same vehicle configuration, it is registered as a vehicle configuration pending verification.

[0044] The sales area is jointly registered by the inventory location, the store location, the customer's origin, and the expected registration location. When the inventory location and the store location are the same, it is registered as a local sales area. When the inventory location and the expected registration location are different and there are transfer records, it is registered as a cross-regional sales area.

[0045] Sales channels are registered as store channels, manufacturer direct sales channels, third-party platform channels, wholesale channels, and auction channels. When the channel field is missing, the source platform number, business document type, and quotation source are read to form a channel pending verification status.

[0046] The observation period consists of historical observation time and future price observation period. The historical observation time is the number of valid days registered for the price caliber version, read backward from the data cutoff time. For new car sales scenarios, the observation period is no less than 30 calendar days and no more than 180 calendar days, and for used car sales scenarios, the observation period is no less than 60 calendar days and no more than 365 calendar days. The future price observation period is registered according to the predicted purpose. For inventory quotation purposes, the observation period is 7 to 30 calendar days, for monthly sales purposes, the observation period is 30 to 90 calendar days, and for inter-regional transfer purposes, the observation period is 15 to 60 calendar days.

[0047] The data cutoff time serves as the time boundary for all input records to enter this session. Records received after the data cutoff time will not be included in the current prediction output version, but will only be registered as candidate records for the next batch.

[0048] The pricing caliber version is used to define the boundaries between the base cash price, the on-the-road price including tax, the total price of the installment plan, the price after trade-in subsidies, the platform listing price, the order locked price, the actual transaction price, the wholesale transaction price, and the inventory clearance price. The sales stage version is used to define the boundaries between the vehicle source filing stage, the platform display stage, the customer inquiry stage, the in-store negotiation stage, the order locked price stage, the delivery and transaction stage, the order cancellation and rollback stage, the wholesale disposal stage, and the clearance disposal stage. The predictive output version is used to register the version number, version creation time, version applicable time, and version call scope of this output.

[0049] Once the session is established, the system reads the vehicle's basic information, inventory, listings, prices, orders, transactions, financing, trade-ins, promotions, vehicle condition, and regional transaction records according to the vehicle object, model configuration, sales region, sales channel, and observation period registered in the session.

[0050] The vehicle basic record reads the vehicle identification number, brand, model series, model year, configuration name, energy type, announcement model, emission standard, range version, and drive type; the inventory record reads the inventory number, entry time, inventory location, inventory age level, preparation status, availability status, transfer status, and inventory batch number; the listing record reads the platform number, vehicle source release number, displayed price, listing time, delisting time, display channel, and platform activity number; the quotation record reads the quotation number, quotation time, quotation channel, quotation amount, sales consultant number, and customer lead number; the order record reads the order number, price lock time, order validity period, deposit status, final payment status, and order status; and the transaction record reads the transaction time. The system retrieves the following information: transaction amount, delivery confirmation status, invoice records, and payment status; financial records include the financial plan number, down payment, number of installments, monthly payment, final payment, financial service fee, and loan interest; trade-in records include the trade-in vehicle appraisal number, used vehicle appraisal amount, trade-in subsidy amount, and deduction method; promotional records include manufacturer price adjustment records, manufacturer cash discount records, dealer promotional activity records, and regional subsidy records; vehicle condition records include the inspection report number, inspection report date, accident disclosure records, maintenance records, mileage records, recall records, and preparation completion time; and regional transaction records include local transaction records, out-of-town transaction records, wholesale transaction records, auction transaction records, logistics cost records, and registration restriction records for the same vehicle model.

[0051] Each input record is registered as a source record when it enters the session. The source record is composed of the data source, receiving batch, vehicle object, record occurrence time, record receiving time, and original serial number. If the original serial number is missing, it is registered as a temporary source record. The temporary source record only enters the verification area and will enter the corresponding price processing flow after the original serial number is supplemented.

[0052] Through the establishment and recording of the above sessions, subsequent steps can continue to process price records under the same vehicle object, same model configuration, same sales region, same sales channel, same observation time, and same version boundary, avoiding the direct mixing of data from different time ranges, different price standards, and different sales stages.

[0053] S2. Based on the source record, price field, expense item, business document status, sales channel, and record occurrence time, generate price source tags, price caliber tags, and sales stage tags. Then, create comparable price groups for each stage based on vehicle type, price caliber, sales stage, regional channel, and price validity period. Specific implementation:

[0054] Read the source record, price field, expense item, business document status, sales channel, and record occurrence time generated by S1, and generate price source label, price caliber label, and sales stage label for each price record. The price source label is formed by the business document type, source platform, price field name, order number, quotation number, financial solution number, replacement record number, payment status, and delivery status in the source record, and is not formed directly by a single field.

[0055] When the source record has a transaction order number, a payment status that is either deposit confirmed, final payment paid, or delivery confirmed, and the vehicle matches the ordered vehicle, the price source label is registered as the actual transaction source; when the source record has a quotation number, sales consultant quotation record, and customer lead number, but no order number has been generated, the price source label is registered as the dealer quotation source; when the source record comes from a third-party vehicle sourcing platform, has a displayed price, vehicle sourcing release number, listing time, and platform number, and no order number has been generated, the price source label is registered as the platform listing source; the source record has... When the source record includes a financing plan number, down payment, monthly payment amount, number of installments, final payment, and financial service fee, the price source label is registered as the source of the financing plan. When the source record includes a trade-in assessment number, used vehicle assessment amount, and trade-in subsidy amount, the price source label is registered as the source of the trade-in subsidy. When the source record comes from manufacturer policy documents and includes the manufacturer's suggested retail price, manufacturer's cash discount amount, and policy effective date, the price source label is registered as the source of manufacturer promotions. When the source record comes from wholesale documents, auction documents, and batch disposal documents, the price source label is registered as the source of wholesale transactions, auction transactions, and inventory clearance, respectively.

[0056] When a business document type is missing, the source platform, price field name, and related number are read to form a business type pending verification status. Price records in the business type pending verification status do not enter the stage comparable price group, but only enter the pending verification area.

[0057] The price caliber label is composed of the price field, expense items, applicable object of the price, and price formation scenario. Expense items must at least include purchase tax, insurance premium, registration fee, interior decoration fee, financial service fee, loan interest, extended warranty fee, delivery fee, logistics fee, preparation fee, trade-in subsidy, manufacturer cash discount, regional subsidy, and dealer concessions.

[0058] When the price field represents the base price of the vehicle, and the expenses do not include purchase tax, insurance, registration fee, financial service fee, and loan interest, the price label is registered as the base cash price. When the price field represents the on-the-road price, and the expenses include purchase tax, insurance, and registration fee, the price label is registered as the on-the-road price including tax. When the price field comes from a financing plan and simultaneously includes down payment, number of installments, monthly payment amount, final payment, and financial service fee, the price label is registered as the total price of the installment plan. When the price field comes from a trade-in record and the price amount has already deducted the trade-in subsidy amount, the price label is registered as the price after the trade-in subsidy. When the price field comes from the platform's display records and no order number or payment status has been generated, the price caliber label is registered as the platform's listed price; when the price field comes from order records and the order is still valid, the price caliber label is registered as the order's locked price; when the price field comes from the transaction record corresponding to the delivery confirmation record, payment record, and invoice record, the price caliber label is registered as the actual transaction price; when the price field comes from wholesale documents, the price caliber label is registered as the wholesale transaction price; when the price field comes from inventory clearance activities and the inventory status meets the clearance disposal conditions, the price caliber label is registered as the inventory clearance price.

[0059] When a fee item is missing, the price record is registered as "fee item pending"; when a fee item includes a financial service fee but the price field is registered as the bare vehicle price, it is registered as "price scope conflict"; when the replacement subsidy amount is deducted but there is no replacement assessment number, it is registered as "replacement scope pending verification".

[0060] Sales stage tags are formed by the status of business documents, the time of record occurrence, order status, inventory status, customer lead status, and sales channel. When a vehicle has only completed inventory filing but has not yet been listed, the sales stage tag is registered as the vehicle filing stage; when a platform vehicle listing number and listing time exist but a customer lead number has not yet been generated, it is registered as the platform display stage; when a customer lead number, inquiry record, and sales consultant quotation record exist but a store visit record has not yet been generated, it is registered as the customer inquiry stage; when a store visit record, test drive record, store quotation, and customer negotiation feedback exist, it is registered as the store negotiation stage; when an order number, price lock time, order validity period, and deposit status exist, it is registered as the order price lock stage; when a delivery confirmation status, final payment status, and invoice record exist, it is registered as the delivery and completion stage; when an order cancellation record, deposit refund record, and price release record exist, it is registered as the order cancellation and rollback stage; when a wholesale batch number and wholesale transaction record exist, it is registered as the wholesale disposal stage; when the inventory status reaches a high inventory age and a clearance activity number exists, it is registered as the clearance disposal stage. If the sales stage field is missing, read the business document status to complete the stage; if the business document status is also missing, register it as a stage pending verification.

[0061] After the price source label, price scope label, and sales stage label are formed, comparable price groups are formed according to vehicle type, price scope, sales stage, regional channel, and price validity period.

[0062] The validity period of a price is determined by the time the record occurs, the time the price becomes effective, the time the price expires, and the validity period of the business document. For order-locked pricing, the order validity period is used as the validity period of the price. For platform-listed prices, the listing time to delisting time is used as the validity period of the price. For promotional prices, the promotion effective time to promotion expiration time is used as the validity period of the price. For actual transaction prices, the transaction time to sample version expiration time is used as the validity period of the price.

[0063] When a price expiration time is missing, the occurrence time of the next price record with the same vehicle, price caliber, and sales stage is read as the temporary expiration time of the previous price record, and the expiration time is recorded as pending verification. When forming a comparable price group for a stage, the platform listing price and the actual transaction price are not placed in the same group, the total price of the installment plan and the cash price of the bare vehicle are not placed in the same group, the wholesale transaction price and the retail transaction price are not placed in the same group, and the price in the cancellation and rollback stage is not used as a valid transaction price.

[0064] This step transforms price records from a state of mixed sources into processable records with source, scope, stage, regional channel, and effective time boundary, providing a unified object for subsequent price event registration and verification by region.

[0065] S3. Based on manufacturer price adjustments, promotional changes, regional subsidies, inventory pressure, competitor price reductions, changes in financial policies, and vehicle condition disclosure records, register price events and their impact windows, and write event impact tags to the corresponding price records. Specific implementation:

[0066] Read the comparable price groups generated by S2 and register price events and their impact windows based on manufacturer price adjustments, promotional changes, regional subsidies, inventory pressure, competitor price reductions, changes in financial policies, and vehicle condition disclosure records. When registering price events, first read the event source, event type, event occurrence time, event reception time, event effective time, event expiration time, affected brand, affected vehicle series, affected model, affected year, affected configuration, affected energy type, affected sales region, affected sales channel, and event credibility status.

[0067] Manufacturer price adjustment events are formed by manufacturer's suggested retail price adjustment records, manufacturer's policy number, vehicle configuration, and the effective time of the price adjustment; promotional change events are formed by manufacturer's cash discount records, dealer promotional activity records, activity number, discount amount, activity effective time, and activity expiration time; regional subsidy events are formed by regional subsidy policy number, subsidy amount, applicable region, applicable energy type, applicable purchaser, and policy effective time; inventory pressure events are formed by inventory region, vehicle configuration, number of units in inventory, inventory days, inventory age level, and clearance activity number; competitor price reduction events are formed by competitor model, competitor sales region, competitor channel, competitor price reduction records, and competitor price effective time; financial policy change events are formed by financial plan number, down payment ratio, number of installments, monthly payment amount, final payment, financial service fee, interest rate change time, and financial institution number; vehicle condition disclosure events are formed by inspection report number, inspection report time, accident disclosure records, maintenance records, recall records, mileage anomaly records, and disclosure receipt time.

[0068] The credibility status of an event is registered according to the source entity and the completeness of the record. Official documents from the manufacturer, manufacturer's back-end policies, publicly disclosed regional subsidy documents, and approved promotional orders are registered as high credibility status; drafts of internal activities by dealers, prices captured by third-party platforms, and notes from sales consultants are registered as pending verification status; events that lack a source entity, lack an effective time of the event, and cannot be matched with price records are registered as low credibility status.

[0069] Once a price event is registered, an impact window is created. The impact window for a manufacturer's price adjustment event begins at the effective time of the price adjustment and ends before the effective time of the next manufacturer's guidance price version. If the next version does not appear, the temporary end time is determined by the number of days of manufacturer price adjustment observation registered in the price event configuration version, and a continuous observation status is recorded.

[0070] The impact window of promotional change events starts from the effective time of the event and ends at the expiration time of the event. If the expiration time of the event is missing, the deadline in the event approval form will be used as the temporary expiration time. If the approval form is also missing, it will be registered as a pending verification status for the promotion expiration time.

[0071] The impact window for regional subsidy events begins from the policy's effective date and ends at the policy's expiration date, and only applies to the regions, energy types, and purchasing entities registered for the subsidy policy. The impact window for inventory pressure events begins when the number of vehicles in stock for the same sales region and the same vehicle model configuration reaches the inventory pressure control value, or when inventory vehicles enter a high-age state, and ends when the number of vehicles in stock falls back to the inventory pressure relief value and remains there for at least two effective sales cycles; an effective sales cycle is a sales day with records of listing, quoting, inquiry, order, or transaction.

[0072] The impact window for competitor price reductions begins from the time the competitor's price is displayed on the main channels in the same region and ends when the price of the target model has not continued to change towards lower prices for at least three consecutive effective sales cycles. The impact window for changes in financial policies begins from the effective date of the new financial plan and ends before the effective date of the next version of the financial plan.

[0073] The impact window for vehicle condition disclosure events begins when inspection reports are updated, repair records are supplemented, accident disclosure records are created, or recall records enter the sales impact scope, and ends when the price records of affected vehicles are adjusted for sample attribution. After the impact window is formed, the vehicle objects, model configurations, sales regions, sales channels, price standards, sales stages, and price occurrence times in the comparable price group are read, and the price records are written to the corresponding event impact markers.

[0074] When a price occurs within the impact window of a manufacturer's price adjustment event, and the vehicle model and configuration are consistent with the scope of the manufacturer's price adjustment event, a manufacturer's price adjustment impact marker is written; when a price occurs within the impact window of a promotional change event, and the sales channel to which the price record belongs participates in a promotional activity, a promotional impact marker is written; when a price occurs within the impact window of a regional subsidy event, and the vehicle's energy type, sales region, and purchaser meet the subsidy application conditions, a regional subsidy impact marker is written; when a price occurs within the impact window of an inventory pressure event, and the vehicle belongs to the corresponding sales region and model configuration, an inventory pressure impact marker is written; when a price occurs within the impact window of a competitor's price reduction event, and the competitor's model is registered as a competitor of the target model in the sales phase version, a competitor's price reduction impact marker is written; when a price occurs within the impact window of a financial policy change event, and the price caliber label is the total price of the installment plan, a financial policy impact marker is written; when a price occurs within the impact window of a vehicle condition disclosure event, and the vehicle is consistent with the inspection report, maintenance record, accident disclosure record, or recall record, a vehicle condition disclosure impact marker is written.

[0075] When the same price record simultaneously meets the requirements of multiple event impact windows, the main event impact flag and the additional event impact flag are registered according to the priority of the price event configuration version registration.

[0076] Vehicle condition disclosure impact markers take precedence over promotion impact markers; manufacturer price adjustment impact markers take precedence over dealer promotion impact markers; inventory pressure impact markers take precedence over platform listing abnormality impact markers; financial policy impact markers only apply to the total price of installment plans, not the cash price of the vehicle.

[0077] If the event occurrence time is missing but the event reception time exists, the event reception time is registered as a temporary event time and an event time pending verification status is formed; if the event reception time is later than the price record reception time and exceeds the lateness allowable time, it is registered as a late event status. Late events only affect the current and subsequent prediction output versions and do not rewrite the historical prediction output versions that have already been called externally.

[0078] Through this step, the price records in the comparable price group of the stage not only have the scope and stage boundaries, but also carry the reasons for the price changes, the effective time of the event, and the scope of application of the event. During subsequent verification, it is possible to distinguish between regular prices, event-affected prices, late event prices, and event prices pending verification.

[0079] S4. Perform source deduplication, cost composition verification, transaction closure verification, vehicle condition verification, inventory status verification, and regional channel verification on comparable price groups for each stage, and write the verification status into the valid sample area, pending verification area, and isolation area. Specific implementation:

[0080] After the S3 write event is marked, perform source deduplication, cost composition verification, transaction closure verification, vehicle condition verification, inventory status verification, and regional channel verification on the comparable price group of the stage. Then write the verification status into the valid sample area, the pending verification area, and the isolation area.

[0081] The source deduplication process first reads the vehicle object, source record, price source tag, price caliber tag, sales stage tag, price amount, record occurrence time, order number, quotation number, and sales channel. If the same vehicle object, order number, price caliber, price amount, and record occurrence time are consistent, the record with the earliest receipt time and complete cost items is registered as the primary source record. The remaining records are registered as duplicate subsidiary records, and these duplicate subsidiary records are written to the duplicate subsidiary status in the pending verification area and do not participate in the formation of the valid sample area.

[0082] When the same vehicle, the same pricing caliber, and the same sales stage are consistent, but the price amount is inconsistent, a price conflict is formed. If the conflict occurs between the platform listing source and the actual transaction source, the platform listing source is written into the display difference status in the pending verification area, while the actual transaction source continues to enter the transaction closed-loop verification. If the conflict occurs between two transaction prices under the same order number, it is written into the transaction price conflict status in the isolation area, waiting for price modification approval, invoice records, refund records, or delivery confirmation records to be supplemented.

[0083] The cost composition verification reads the price caliber label and cost items. If the record of the bare vehicle cash price shows that purchase tax, insurance premium, registration fee, financial service fee, loan interest, or trade-in subsidy have been deducted but not separated, write the cost item status as "pending verification" in the verification area. If the record of the tax-inclusive on-the-road price is missing any of the purchase tax, insurance premium, or registration fee, write the on-the-road cost status as "pending verification" in the verification area. If the record of the installment plan total price is missing any of the fields for down payment, number of installments, monthly payment amount, final payment, or financial service fee, write the financial field status as "pending verification" in the verification area. If the record of the trade-in subsidy price is missing any of the fields for trade-in assessment number, old vehicle assessment amount, or subsidy source number, write the trade-in caliber status as "pending verification" in the verification area.

[0084] Records with complete expense items and price labels consistent with the expense items continue to the transaction closure verification process. The transaction closure verification process reads the order number, vehicle type, transaction amount, deposit status, final payment status, delivery confirmation status, invoice records, and cancellation records.

[0085] When an order lock-in record is within the order validity period and the deposit status is confirmed, it is written into the order lock-in sample in the valid sample area; when an actual transaction record has an order number, final payment status, delivery confirmation status, and invoice record, it is written into the actual transaction sample in the valid sample area; when a transaction record lacks an invoice record but the final payment has been made and delivery has been confirmed, it is written into the invoice pending status in the pending verification area; when a transaction record has not been paid for the final payment and delivery has not been confirmed, it is written into the transaction pending confirmation status in the pending verification area; when a transaction record corresponds to a cancellation record, deposit refund record, and price release record, it is written into the cancellation rollback status in the isolation area and does not enter the valid sample area.

[0086] The vehicle condition verification system reads the inspection report number, inspection report date, accident disclosure records, maintenance records, mileage records, recall records, and preparation completion time.

[0087] If the inspection report number exists, the inspection report time is within the validity period of the vehicle condition configuration version registration, and the accident disclosure record, mileage record, and recall record have not formed a major pending verification status, the vehicle condition is registered as a reliable status; if the inspection report is missing, the inspection report is expired, the mileage reading is abnormal, the accident disclosure record is late, or the recall record is not processed, the vehicle condition is registered as pending verification status.

[0088] Price records in the pending vehicle condition status are placed in the pending verification area and are not considered valid transaction samples for a single vehicle. If there are insufficient valid samples of the same vehicle model and the price output purpose allows for backup samples, the pending vehicle condition records are placed in the backup sample area and are not placed in the valid sample area. After a vehicle condition disclosure event occurs, the transaction prices before the disclosure are written into the pre-disclosure sample status in the isolation area in subsequent prediction output versions, and can only be read when the output purpose is historical reference.

[0089] The inventory status verification reads the inbound time, inventory days, inventory age level, preparation status, sellable status, transfer status, and clearance activity number. The inventory days are recorded according to the inbound time to the data cutoff time. If the inbound time is missing, the initial inventory filing time is used as the temporary inbound time, and the inbound time is recorded as the pending verification status in the pending verification area.

[0090] Inventory days less than 30 days are registered as newly received; 31 to 90 days are registered as regular inventory; 91 to 180 days are registered as high-age inventory; and over 180 days are registered as clearance pending control. Enterprises can register different day boundaries in the inventory status configuration version, but only one fixed boundary is used in the same forecast output version. Inventory clearance prices are only included in the valid sample area for clearance samples when the inventory status is high-age or clearance pending control, and a clearance activity number exists. Low-price records generated under conditions of incomplete preparation, incomplete documentation, pending recall, and incomplete transfer are written to the inventory anomaly status in the isolation area.

[0091] The regional channel verification reads the inventory location, store location, customer origin, expected registration location, sales channel, cross-regional transfer records, logistics cost records, and regional subsidy application conditions.

[0092] When the inventory location, store location, and expected registration location are the same, it is registered as a local channel; when the inventory location and expected registration location are different and there is a transfer approval record, it is registered as a cross-regional channel; when there is no transfer approval record but the price record uses the price of another region, it is written as a regional channel pending verification status in the pending verification area.

[0093] Regional subsidies are only recorded in the corresponding price record when the sales region, energy type, purchasing entity, and policy validity period are all met. If any condition is missing, the subsidy eligibility status is written to the pending verification area.

[0094] After the above verification, price records that are complete in scope, clearly defined in stage, valid in time, complete in transaction loop, reliable in vehicle condition, available in inventory, and consistent with regional channels are written into the valid sample area; records of pending fees, pending business type verification, pending stage verification, pending invoice verification, pending vehicle condition verification, pending warehousing time verification, pending regional channel verification, and pending subsidy eligibility verification are written into the pending verification area; records of price conflicts, transaction price conflicts, pending target verification, order cancellations and rollbacks, abnormal inventory, samples before vehicle condition disclosure, and invalid source records are written into the isolation area.

[0095] The valid sample area, the area awaiting verification, and the isolation area all record the sample entry time, sample source record, reason for status formation, and corresponding session number. When subsequent data is supplemented, the status is updated according to the record number of the same source to avoid samples moving between different areas without a basis.

[0096] S5. Based on the valid sample area, event impact markers, vehicle condition reliability, inventory status, and regional channel status, generate a vehicle sales price forecast version corresponding to the forecast purpose, price caliber, sales stage, and future price observation period. Specific implementation:

[0097] Read the valid sample area, the area to be verified, and the isolation area formed by S4, and generate a car sales price forecast version corresponding to the forecast purpose, price caliber, sales stage, and future price observation period based on the valid sample area, event impact marker, vehicle condition credibility status, inventory status, and regional channel status.

[0098] The forecast usage should at least include inventory quotation usage, customer inquiry usage, order price locking usage, clearance disposal usage, wholesale disposal usage, and inter-regional transfer usage. For each forecast usage, the corresponding price caliber and sales stage should be read, and price records should not be retrieved across different calibers.

[0099] The inventory quotation function retrieves the bare vehicle cash price, platform display stage, customer inquiry stage, and inventory status. The customer inquiry function retrieves the corresponding sales channel, customer origin, pricing caliber, and inquiry time. The order lock-in function retrieves the order lock-in stage, order validity period, and payment status. The clearance sale function retrieves the inventory clearance price, clearance sale stage, inventory pressure impact marker, and clearance activity number. The wholesale sale function retrieves the wholesale transaction price, wholesale sale stage, and wholesale batch number. The cross-regional transfer function retrieves the transfer-out region, transfer-in region, logistics cost records, regional subsidy application status, and cross-regional channel status.

[0100] When a forecast version is generated, a forecast version record is first established based on the vehicle type, forecast purpose, price caliber, sales stage, sales region, sales channel, and future price observation period. Then, the valid sample area under the same boundary is read. If there are actual transaction samples in the valid sample area, the actual transaction samples form the transaction benchmark sample layer. If there are insufficient actual transaction samples and the order lock-in price samples are within the order validity period, the order lock-in price samples form the lock-in price benchmark sample layer. If both the transaction benchmark sample layer and the lock-in price benchmark sample layer are insufficient, the dealer quotation samples and platform listing samples form the reference benchmark sample layer, and the insufficient transaction sample status is recorded in the forecast version.

[0101] For clearance sales, only clearance samples and high-inventory-age transaction samples are read, and the price of newly entered inventory is not read; for wholesale sales, only the source of wholesale transactions and auction transactions are read, and the source of retail transactions is not read; for installment plan total price, only the source of the financing plan is read, and the cash price of the bare vehicle is not read; for trade-in subsidy price, only samples with trade-in assessment number and subsidy source number are read, and the ordinary transaction price is not read directly.

[0102] When an event impact marker is included in the forecast version, samples are categorized into pre-event, during-event, and post-event samples based on the event impact window. Manufacturer price adjustments, regional subsidies, financial policies, and vehicle condition disclosures are classified as strong impact markers. Once a strong impact marker is formed, pre-event samples are not included in the current valid forecast version but are only registered as historical references. Promotional impacts, inventory pressure impacts, and competitor price reduction impacts are included in the event impact sample layer according to the effective scope registered in the price event configuration version.

[0103] When a vehicle condition is considered reliable and enters the prediction version, reliable vehicle condition samples are given priority to enter the single-vehicle prediction version. Vehicle condition samples awaiting verification cannot be used as valid single-vehicle samples. They can only enter the backup sample layer when there are insufficient vehicle model samples and the prediction purpose allows it, and a vehicle condition awaiting verification prompt is registered in the prediction version.

[0104] When inventory status enters the forecast version, new arrivals and regular inventory status enter the regular pricing version, while high inventory age and clearance pending control status enter the inventory pressure impact version; abnormal inventory status does not enter the effective forecast version. When regional channel status enters the forecast version, local channel status generates a local price forecast version, cross-regional channel status generates a cross-regional price forecast version, and regional channel pending verification status does not generate an effective price, only a regional channel pending verification version.

[0105] The price range output by the predicted version is formed from samples within the valid sample area that are of the same vehicle type, price caliber, sales stage, sales region, sales channel, and future price observation period. If the sample size does not reach the minimum sample size registered for the price caliber version, no valid price version is generated, and it is registered as insufficient sample. The predicted version status must be registered at least as follows: pending activation, valid, under event impact, under observation, anomaly isolated, rolled back to activation, and no valid price.

[0106] When the status is pending, read the number of valid samples, the status of the isolation zone, the status of the pending verification zone, the status of the event impact window, the vehicle condition reliability status, the inventory status, and the regional channel status. When the number of valid samples reaches the minimum number of samples, there are no transaction price conflicts or pending verification statuses in the isolation zone, there are no pending fee items or pending vehicle condition verification statuses in the pending verification zone, the event impact window has formed a clear effective range, the inventory status is available for sale, and the regional channels are consistent, the prediction version status is registered as valid.

[0107] When the event impact window is still active and samples are available, the prediction version status is registered as "Event Impacting," and the event type and impact window are registered simultaneously. When there are insufficient valid samples but the backup sample layer is available, the prediction version status is registered as "Observation Pending Verification."

[0108] When there are conflicting transaction prices, pending verification of objects, abnormal inventory, or unprocessed samples before vehicle condition disclosure in the isolation zone, the forecast version status is registered as abnormal isolation. If the current forecast version cannot take effect and the previous valid forecast version is still within its applicable time range, and has not been negated by manufacturer price adjustments, regional subsidies, or vehicle condition disclosure, the previous valid forecast version is read and registered as rolled back to take effect; if the previous valid forecast version has exceeded its applicable time range, or has been marked as negated by a strong influence, rollback is not performed, and the status is registered as no valid price.

[0109] Once the forecast version is generated, register the forecast output version, sample version number, price caliber, sales stage, forecast purpose, future price observation period, event impact marker, vehicle condition reliability status, inventory status, regional channel status, version status, and prohibition notice.

[0110] The following statements are prohibited: The total price of the installment plan must be registered at least; the wholesale price must not be used as the retail price; the price pending vehicle condition verification must not be used as the single-vehicle transaction price; the price affected by the event must not be used as a long-term fixed price; and the price pending verification from regional channels must not be used for cross-regional quotations.

[0111] When new data is added, it does not directly overwrite the already effective prediction version. Instead, it first forms a new sample batch. The new sample batch is then processed through the same process from S2 to S4 to generate a new candidate prediction version. The candidate prediction version replaces the current effective version only after it passes the version status verification. If it fails, the current effective version is retained and the reason for the failure is recorded.

[0112] Through this process, a fixed correspondence is established between the predicted version of automobile sales prices and the prediction purpose, price scope, sales stage, future price observation period, and sample version. The price output is no longer an isolated price value, but a prediction version with source boundaries, scope boundaries, event boundaries, sample boundaries, and applicable boundaries.

[0113] In this embodiment, the minimum number of valid samples registered for the price caliber version is 5, and the number of samples to be observed and verified is 3 to 4. The price conflict control value is 3% of the most recent valid transaction price of the same model. If the price difference between two sources reaches the price conflict control value under the same vehicle object, the same price caliber, the same sales stage, and the same regional channel, and cannot be explained by expense items, financial plans, trade-in subsidies, or promotional activities, it will be written into the isolation area.

[0114] Inventory status is determined from the date of receipt to the data cutoff time. Receipts within 30 days are considered new receipts; 31 to 90 days are considered regular inventory; 91 to 180 days are considered high-age inventory; and over 180 days are considered clearance pending. Inspection reports support a reliable vehicle condition status for 30 days from the report date. If no new inspection report is received after 30 days, the vehicle condition is entered into a pending verification status. Records occurring more than three calendar days after receipt are registered as late records. Late records do not rewrite already publicly available prediction output versions; they only enter the pending verification process for the current and subsequent prediction output versions.

[0115] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0116] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0117] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting automobile sales prices based on multi-source data fusion, characterized in that, include: S1. Establish a car sales price forecasting session, register vehicle objects, model configurations, sales regions, sales channels, observation time, data cutoff time, price caliber version, sales stage version, and forecast output version, and read vehicle basics, inventory, listings, quotations, orders, transactions, financing, trade-ins, promotions, vehicle condition, and regional transaction records. S2. Based on the source record, price field, expense item, business document status, sales channel and record occurrence time, form price source label, price caliber label and sales stage label, and form stage comparable price groups according to vehicle object, price caliber, sales stage, regional channel and price validity period; S3. Based on manufacturer price adjustments, promotional changes, regional subsidies, inventory pressure, competitor price reductions, changes in financial policies, and vehicle condition disclosure records, register price events and their impact windows, and write event impact markers to the corresponding price records. S4. Perform source deduplication, cost composition verification, transaction closure verification, vehicle condition verification, inventory status verification, and regional channel verification on the comparable price group of the stage, and write the verification status into the valid sample area, the pending verification area, and the isolation area. S5. Based on the valid sample area, event impact markers, vehicle condition reliability, inventory status, and regional channel status, generate a version of the automobile sales price forecast that corresponds to the forecast purpose, price caliber, sales stage, and future price observation period.

2. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S1: Vehicle objects are formed by registering individual vehicle objects, vehicle model objects, and regional vehicle model objects. Individual vehicle objects are jointly defined by vehicle identification number, internal inventory number, batch number upon entry, and store inventory number. Vehicle model objects are jointly defined by brand, series, model, year, configuration name, energy type, and drive type. Regional vehicle model objects are based on vehicle model objects and include sales region and sales channel.

3. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S1: The observation period consists of historical observation time and future price observation period. The historical observation time is read backward from the data cutoff time. The future price observation period is registered according to the predicted purpose, which includes the corresponding purpose in inventory quotation, customer inquiry, order price locking, clearance disposal, wholesale disposal and cross-regional transfer.

4. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S2: The source record is formed by the data source, receiving batch, vehicle object, record occurrence time, record receiving time and original serial number. When the original serial number is missing, a temporary source record is formed. The temporary source record enters the verification area, and after the original serial number is supplemented, the process of forming price source label, price caliber label and sales stage label continues according to the supplemented source record.

5. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S2: Price source tags are formed based on business document type, source platform, price field name, order number, quotation number, financial plan number, trade-in record number, payment status, and delivery status. Actual transaction source is jointly registered by transaction order number, payment status, and delivery status. Dealer quotation source is jointly registered by quotation number, sales consultant quotation record, and customer lead number. Platform listing source is jointly registered by vehicle source release number, display price, listing time, and platform number.

6. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S2: The price caliber label is formed based on the price field and the expense items. The expense items include purchase tax, insurance premium, license plate fee, decoration fee, financial service fee, loan interest, extended warranty fee, delivery fee, logistics fee, preparation fee, trade-in subsidy, manufacturer cash discount, regional subsidy and dealer discount. When an expense item is missing, the corresponding price record is recorded as expense item pending.

7. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S2: Sales stage tags are formed based on the status of business documents, the time of record occurrence, order status, inventory status, customer lead status, and sales channel. The sales stage tags are one of the following: vehicle source filing stage, platform display stage, customer inquiry stage, in-store negotiation stage, order price locking stage, delivery and transaction stage, order cancellation and rollback stage, wholesale disposal stage, and clearance disposal stage. When both the sales stage field and the business document status are missing, the corresponding price record is registered as a stage pending verification status.

8. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S3: The impact window of a price event is formed according to the type of event. The impact window of a manufacturer's price adjustment event starts from the effective time of the price adjustment and ends before the effective time of the next manufacturer's suggested retail price version. The impact window of a promotional change event starts from the effective time of the promotion and ends at the expiration time of the promotion. The impact window of an inventory pressure event starts when the number of inventory units reaches the inventory pressure control value or when the inventory vehicles enter a high inventory age state and ends when the inventory pressure relief value continuously meets two effective sales cycles.

9. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S4: The valid sample area accepts price records that are complete in terms of price scope, clear in terms of sales stage, have not expired in terms of price validity period, complete in terms of transaction loop, reliable in terms of vehicle condition, have available inventory and are consistent in terms of regional channels. Price records for the pending verification area, including items for receiving fees pending verification, business types pending verification, stages pending verification, invoices pending verification, vehicle conditions pending verification, warehousing time pending verification, regional channels pending verification, and subsidy eligibility pending verification; The quarantine zone receives price records for price conflicts, transaction price conflicts, pending verification of objects, order cancellations and rollbacks, inventory anomalies, samples before vehicle condition disclosure, and invalid sources.

10. The method for predicting automobile sales prices based on multi-source data fusion according to claim 1, characterized in that, In S5: The automobile sales price forecast version registration includes the forecast output version, sample version number, price caliber, sales stage, forecast purpose, future price observation period, event impact marker, vehicle condition reliability status, inventory status, regional channel status, and version status. The version status is one of the following: pending activation, effective, affected by the event, under observation pending verification, isolated from an anomaly, rolled back to effective, or no effective price.

Images