Key customer vehicle maintenance monitoring system

Abstract

The invention belongs to the field of customer management, and discloses a key customer vehicle maintenance monitoring system, which comprises a prediction module, and the prediction module comprises a weight updating unit, a prediction unit and a probability calculation unit. The weight updating unit is used for calculating the weighted weight of each model used for prediction, and the weight updating unit comprises the steps that the parking probability of each model used for prediction in the latest N times of prediction is obtained, and N is a self-adaptive parameter; calculating a weighted weight based on the parking probabilities of the latest N times of prediction and a set probability threshold value; the prediction unit is used for predicting the parking probability of the key customer by using each model for prediction to obtain the parking probability; and the probability calculation unit is used for weighting all the parking probabilities based on the weighted weights to obtain weighted parking probabilities. According to the invention, the accuracy of the prediction result can be further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of customer management, in particular to a key customer vehicle maintenance monitoring system. BACKGROUND

[0002] In order to provide a better parking experience for key customers (such as leader vehicles and senior member vehicles), in the prior art, it is determined whether to reserve a parking space in advance by predicting the probability of the key customer parking in a specified time period in the future. In the existing prediction method, in order to improve the stability of the prediction result, the predicted parking probabilities of multiple models are usually considered to obtain the final predicted parking probability by weighting, and the weights of the weighting process are usually the same, which also leads to that the model with a lower prediction accuracy in a short period obtains an excessively high weight, resulting in that the accuracy of the final prediction result is not high enough. SUMMARY

[0003] The purpose of the present application is to disclose a key customer vehicle maintenance monitoring system to solve the technical problems pointed out in the background.

[0004] In order to achieve the above purpose, the present application adopts the following technical solutions: The present application provides a key customer vehicle maintenance monitoring system, comprising a prediction module, the prediction module comprising a weight updating unit, a prediction unit and a probability calculation unit; The weight updating unit is used to calculate the weighting weight of each model used for prediction, comprising: Respectively obtaining the parking probability of each model used for prediction in the last N times of prediction, N being an adaptive parameter; Calculating the weighting weight based on the parking probability of the last N times of prediction and the set probability threshold; The prediction unit is used to predict the parking probability of the key customer using each model used for prediction respectively, and obtain the parking probability; The probability calculation unit is used to weight all the parking probabilities based on the weighting weight, and obtain the weighted parking probability.

[0005] Preferably, the determination process of the adaptive parameter is as follows: Obtaining the standard value M of the set adaptive parameter; After each prediction ends, the adaptive parameter N used in the next prediction process is calculated based on the weighted parking probability obtained by the last M times of prediction and the set probability threshold.

[0006] Preferably, the adaptive parameter N used in the next prediction process is calculated based on the weighted parking probability obtained by the last M times of prediction and the set probability threshold, comprising: obtaining a number num1 of weighted parking probabilities greater than a set probability threshold from the weighted parking probabilities obtained in the last M predictions; obtaining a number num2 of weighted parking probabilities greater than a set probability threshold from the weighted parking probabilities obtained in the last M predictions; obtaining a number num2 of weighted parking probabilities greater than a set probability threshold from the weighted parking probabilities obtained in the last M predictions; denotes rounding up; calculating an adaptive parameter N based on num1 and num2.

[0007] Preferably, the weighted weight is calculated based on the parking probabilities of the last N predictions and the set probability threshold, comprising: storing all models used for prediction into a set U; obtaining a number of parking probabilities greater than a set probability threshold from the parking probabilities of the last N predictions for each model in the set U respectively; converting the parking probabilities of the last N predictions for each model in U into a result sequence respectively based on the set probability threshold; calculating a first weight for each model in U based on the result sequence respectively; calculating a second weight for each model in U based on the number of parking probabilities greater than the set probability threshold respectively; calculating the weighted weight based on the first weight and the second weight.

[0008] Preferably, converting the parking probabilities of the last N predictions for each model in U into a result sequence respectively comprises: for a model q in U, storing the parking probabilities of the last N predictions for model q in a sequence in order of prediction time from early to late; processing each parking probability in the sequence to obtain a result sequence respectively: for the i-th parking probability in , if is greater than the set probability threshold, modifying the value of to 1, otherwise modifying the value of to 0.

[0009] Preferably, further comprising a model training module; the model training module is configured to train each model in the set V to obtain a model used for prediction; V is a set of untrained models with parking probability prediction capability.

[0010] Preferably, the models in the set V include gradient boosting tree models, naive Bayes models, LSTM neural network models, and Bayesian structure time series models.

[0011] Preferably, the system further comprises a training data collection module. The training data collection module is configured to obtain data for training the model.

[0012] Preferably, the data for training the model comprises: customer attribute data, time feature data and external environment data.

[0013] Preferably, the system further comprises a prompt module. The prompt module is configured to prompt the parking lot manager based on the weighted parking probability.

[0014] Preferably, the system further comprises a license plate monitoring module. The license plate monitoring module is configured to obtain the license plate number of a vehicle entering the parking lot and determine whether the vehicle owner is a VIP based on the license plate number.

[0015] Beneficial effects: In the process of predicting the parking probability of the VIP by using multiple models, the present application does not use equal weights to weight the prediction results of various models, but calculates the weighting weight based on the parking probability obtained by the recent N times of prediction and the set probability threshold, so that the weighting weight can change with the change of the prediction result. By comparing the prediction results of different models, the model that predicts more accurately in the recent period can have a greater impact on the final weighted parking probability, thereby further improving the accuracy of the prediction result. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0017] Figure 1 The first schematic diagram of the VIP vehicle maintenance monitoring system of the present application.

[0018] Figure 2 The schematic diagram of the process of calculating the weighting weight.

[0019] Figure 3 The second schematic diagram of the VIP vehicle maintenance monitoring system of the present application.

[0020] Figure 4 The third schematic diagram of the VIP vehicle maintenance monitoring system of the present application.

[0021] Figure 5It is a fourth schematic view of the vehicle maintenance monitoring system for the key customer of the present application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0023] As shown in an embodiment, Figure 1 The vehicle maintenance monitoring system for the key customer provided by the present application comprises a prediction module, and the prediction module comprises a weight updating unit, a prediction unit and a probability calculation unit. The weight updating unit is used for calculating the weighted weight of each model used for prediction, comprising: The parking probability of each model used for prediction in the last N times of prediction is obtained respectively, and N is an adaptive parameter; The weighted weight is calculated based on the parking probability of the last N times of prediction and the set probability threshold value; The prediction unit is used for predicting the parking probability of the key customer by using each model used for prediction respectively, and obtaining the parking probability. The probability calculation unit is used for weighting all the parking probabilities based on the weighted weight, and obtaining the weighted parking probability.

[0024] In the present application, the predicted parking probability of the model used for prediction is normalized to the interval [0, 1].

[0025] The set probability threshold value of the present application can be 0.7. Further, the set probability threshold value can be adjusted according to the prediction result and the parking result, for example, if it is found that a lower set probability threshold value can have a higher success rate to judge whether the key customer parks or not in the last time period (for example, one month), the set probability threshold value is correspondingly reduced in the next time period.

[0026] In the present application, the parking probability of each time period is usually predicted respectively, for example, 9:00 to 12:00 is a time period, and 14:00 to 18:00 is a time period. Therefore, in the process of training the prediction model, the same model needs to be trained for each time period to improve the pertinence of the model and the accuracy of the prediction result.

[0027] Preferably, the adaptive parameter determination process is as follows: Obtain the standard value M of the set adaptive parameter; After each prediction, calculate the adaptive parameter N used in the next prediction process based on the weighted parking probability obtained in the last M predictions and the set probability threshold.

[0028] The standard value M of the present application can be 20. That is, the adaptive parameter N is calculated by obtaining the prediction results of the last 20 times.

[0029] In the initial running stage, if the cumulative prediction times are less than M, the value of N is directly set to M.

[0030] Preferably, the adaptive parameter N used in the next prediction process is calculated based on the weighted parking probability obtained in the last M predictions and the set probability threshold, including: Obtain the number num1 of weighted parking probabilities greater than the set probability threshold in the weighted parking probabilities obtained in the last M predictions; Obtain the number num2 of weighted parking probabilities greater than the set probability threshold in the weighted parking probabilities obtained in the last M predictions; num2=M-num1; represents the upward rounding; Calculate the adaptive parameter N based on num1 and num2, including: First step, calculate the number num3: num3=num1-num2; Second step, calculate the value of N: .

[0031] In the process of calculating the adaptive parameter N, the present application divides the weighted parking probabilities greater than the set probability threshold in the last M predictions into two intervals, i.e. +1 to M is one interval, and 1 to is the other interval. The calculation is performed by comparing the changes between the weighted parking probabilities in the two intervals, so that the value of N of the present application can adaptively change with the change of the weighted parking probability, so that the prediction accuracy can be improved, and the higher the accuracy is, the smaller the value of N is, so as to improve the calculation efficiency of the weighted weight; otherwise, the value of N is larger, and the weighted weight can more accurately represent the prediction accuracy of the model based on more historical data.

[0032] Preferably, as Figure 2 the weighted weight is calculated based on the parking probability of the last N predictions and the set probability threshold, including: Firstly, all models used for prediction are stored in a set U.

[0033] Secondly, the number of parking probabilities greater than the set probability threshold in the last N predictions of each model in the set U is obtained respectively.

[0034] Thirdly, based on the set probability threshold, the parking probabilities of each model in U in the last N predictions are converted into result sequences respectively, including: For model q in U, the parking probabilities of model q in the last N predictions are stored in sequence in order of prediction time from early to late. Each parking probability in sequence is processed as follows to obtain the result sequence: For the i-th parking probability in , if is greater than the set probability threshold, the value of is modified to 1, otherwise the value of is modified to 0.

[0035] For example, if the parking probabilities of the last 5 predictions are 0.5, 0.8, 0.9, 0.6 and 0.3, then sequence = [0.5, 0.8, 0.9, 0.6, 0.3]. When the set probability threshold is 0.7, the result sequence is [0, 1, 1, 0, 0].

[0036] By converting the sequence of parking probabilities into the result sequence, the differences in the prediction results of different models are amplified compared to directly using the parking probabilities to calculate the first weight, which can more effectively represent the differences in the prediction results between various models in the subsequent calculation process, thereby effectively increasing the weighted weight of the model with stronger prediction result independence under the same prediction success rate, and improving the effectiveness of the final weighted parking probability. When the prediction success rates of different models meet the use requirements (for example, the prediction success rate is greater than 0.8), if the correlation between the parking probability sequence of a model and the parking probability sequences corresponding to other models is lower, it indicates that the prediction result is more independent.

[0037] Fourthly, the first weight of each model in U is calculated based on the result sequence, including: S1, for model q in U, the difference coefficient between the prediction results of q and each model in U except model q is calculated based on the result sequence respectively:

[0038] is the difference coefficient of the prediction results between model q and model r; is the judgment value, if The kth element in If the kth elements in are different, then is 1, otherwise 0; is the time length between the generation time of the parking probability corresponding to the kth element and the current time, is the weight control factor of the kth element; The acquisition process is as follows: The standard deviation of the parking probabilities corresponding to the kDth to kth elements in the calculation result sequence is calculated; D is a preset integer, for example, D can be 5.

[0039] When k is less than D, the result of kD is set to k, which can avoid kD being negative due to insufficient number of previous elements.

[0040] Normalize the standard deviation and map it to the interval [0,1] to obtain the standardized standard deviation ; Standard deviation calculation based on normalization :

[0041] and are the maximum and minimum values ​​of the weight control factors corresponding to the first k elements in the result sequence respectively; S2, calculate the first weight of model q using the following formula:

[0042] Uq represents the set of all models in U except model q; is the first weight of model q.

[0043] The first weight of the present invention is calculated based on the coefficient of difference. Therefore, the larger the sum of the coefficients of difference between model q and other models, the larger the first weight of model q, indicating that the independence between the prediction results of model q and the prediction results of other models is stronger. In addition, in the process of calculating the coefficient of difference, the present invention also sets a corresponding influence coefficient for each judgment value according to the position of the elements in the model in the result sequence. Therefore, if the generation time of the parking probability corresponding to the element in the result sequence is farther away from the current time, the greater the volatility of the parking probability corresponding to the latest D elements, the smaller the influence of the judgment value corresponding to the element on the first weight, thereby improving the response sensitivity of the weighted weight of the present invention to changes in the prediction accuracy of the model.

[0044] When the near-term parking probability exhibits high volatility (high variance), this may indicate unstable or noisy forecasts. In this case, the influence coefficient should be slowed down, giving relatively higher weight to more distant, potentially more stable data to smooth out the noise. Conversely, when data volatility is low (stable), the influence coefficient can be accelerated, placing greater emphasis on recent trends. This can more promptly enhance the impact of models with increasing prediction accuracy on the final weighted results, thereby improving the accuracy of the final forecast.

[0045] The parking probability corresponding to the elements in the result sequence can be obtained from the sequence based on the position of the element Obtained from.

[0046] For example, for model q, The sequence is obtained by sorting the parking probabilities predicted by model q in the last N times in order of prediction time from early to late; Then the parking probability corresponding to the kth element in the result sequence is The kth element in .

[0047] The fifth step is to calculate the second weight of each model in U based on the number of parking probabilities greater than the set probability threshold, including: For model q, the number of parking probabilities predicted by model q in the last N times that are greater than the set probability threshold is expressed as ; The second weight of model q is calculated using the following formula:

[0048] is the second weight of model q.

[0049] The sixth step is to calculate the weighted weight based on the first weight and the second weight, including: The weighted weight is calculated using the following formula:

[0050] is the weighted weight of model q.

[0051] The weighted weight of the present application not only considers the independence of the prediction result of the model, but also considers the accuracy of the prediction result of the model, so that if the accuracy of the prediction result of the model is higher and the independence is stronger, the weighted weight is larger, so that the final prediction accuracy after the prediction results of each model are integrated is effectively improved, and the stability of the final prediction result is improved.

[0052] Preferably, the parking probabilities of all are weighted based on the weighted weights to obtain weighted parking probabilities, comprising:

[0053] For the weighted parking probability, is the parking probability output by the model q in the latest prediction.

[0054] Preferably, as Figure 3 The present application further comprises a model training module. The model training module is used for training each model in the set V to obtain a model for prediction. V is a set of untrained models with parking probability prediction capability.

[0055] During training, the training data can be divided into different independent sets (for example, into a training set and a test set), and the training of the model and the evaluation of the prediction accuracy of the model are carried out based on different sets.

[0056] Preferably, the models in the set V include gradient boosting tree models, naive Bayes models, LSTM neural network models and Bayesian structure time series models.

[0057] Preferably, as Figure 4 The present application further comprises a training data collection module. The training data collection module is used for obtaining data for training the model.

[0058] Specifically, the training data collection module can obtain data from the database of the parking lot management system and data from the network.

[0059] Preferably, the data for training the model comprises: Customer attribute data, time feature data and external environment data.

[0060] The customer attribute data includes membership level, historical parking frequency, age, etc. The time feature data includes which day of the week the parking time belongs to, which time period in a day the parking time belongs to, etc. The external environment includes the type of outdoor weather when the key customer parks, the number of large-scale activities around the parking lot, etc.

[0061] Preferably, as Figure 5 The application also comprises a prompting module. The prompting module is used to prompt the parking lot manager based on the weighted parking probability.

[0062] Specifically, when the weighted parking probability is greater than the set probability threshold, the license plate number of the vehicle of the VIP, the name of the VIP, and other information are sent to the device used by the parking lot manager, so that the parking lot manager can reserve a parking space for the VIP in advance.

[0063] Preferably, it also comprises a license plate monitoring module. The license plate monitoring module is used to obtain the license plate number of the vehicle entering the parking lot, and determine whether the vehicle owner is a VIP based on the license plate number.

[0064] The prompting module is also used to prompt the parking lot manager when the license plate monitoring module determines that the vehicle owner is a VIP based on the license plate number. In this way, the parking lot manager can arrange parking services for the VIP in advance according to the parking habits of the VIP.

[0065] The above is only a specific implementation of the application, but the protection scope of the application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the application, which should be covered within the protection scope of the application. Therefore, the protection scope of the application should be subject to the protection scope of the claims.

Claims

1. Key customer vehicle maintenance monitoring system, characterized by: The prediction module includes a weight updating unit, a prediction unit and a probability calculation unit; The weight update unit is used to calculate the weighted weights of each model used for prediction, including: Obtain the parking probability of each prediction model in the last N predictions, where N is an adaptive parameter; Calculate the weighted weight based on the parking probability of the last N predictions and the set probability threshold; The prediction unit is used to predict the parking probability of key customers using each prediction model to obtain the parking probability; The probability calculation unit is used to weight all parking probabilities based on weighted weights to obtain a weighted parking probability.

2. The key customer vehicle maintenance monitoring system according to claim 1 is characterized in that: The process of determining the adaptive parameters is as follows: Get the standard value M of the set adaptive parameter; After each prediction is completed, the adaptive parameter N used in the next prediction process is calculated based on the weighted parking probability obtained from the most recent M predictions and the set probability threshold.

3. The key customer vehicle maintenance monitoring system according to claim 2 is characterized in that: The adaptive parameter N used in the next prediction process is calculated based on the weighted parking probability obtained from the most recent M predictions and the set probability threshold, including: Obtain the number num1 of weighted parking probabilities obtained from the most recent M predictions that are greater than a set probability threshold; Get the latest The number num2 of weighted parking probabilities obtained from the prediction that are greater than the set probability threshold; Indicates rounding up; Calculate the adaptive parameter N based on num1 and num2.

4. The key customer vehicle maintenance monitoring system according to claim 1, characterized in that: The weighted weight is calculated based on the parking probability of the last N predictions and the set probability threshold, including: Store all models used for prediction into set U; Obtain the number of parking probabilities greater than the set probability threshold among the most recent N predicted parking probabilities for each model in the set U; Based on the set probability threshold, the parking probability predicted by each model in U in the last N times is converted into a result sequence; Calculate the first weight of each model in U based on the result sequence; Calculate a second weight for each model in U based on the number of parking probabilities greater than a set probability threshold; A weighted weight is calculated based on the first weight and the second weight.

5. The key customer vehicle maintenance monitoring system according to claim 4 is characterized in that: Convert the parking probability predicted by each model in U in the last N times into a result sequence, including: For model q in U, store the parking probability predicted by model q in the last N times in the sequence from early to late according to the prediction time ; Sequence Each parking probability in is processed as follows to obtain the result sequence: for The i-th parking probability in ,like If the probability is greater than the set threshold, Change the value of to 1, otherwise Change the value to 0.

6. The key customer vehicle maintenance monitoring system according to claim 1, characterized in that: It also includes a model training module; The model training module is used to train each model in the set V to obtain a model for prediction; V is a collection of untrained models with parking probability prediction capabilities.

7. The key customer vehicle maintenance monitoring system according to claim 6, characterized in that: The models in set V include gradient boosting tree model, naive Bayes model, LSTM neural network model and Bayesian structured time series model.

8. The key customer vehicle maintenance and monitoring system according to claim 1, characterized in that: It also includes a training data collection module; The training data collection module is used to obtain data for model training.

9. The key customer vehicle maintenance monitoring system according to claim 8, characterized in that: The data used to train the model includes: Customer attribute data, time characteristic data and external environment data.

10. The key customer vehicle maintenance and monitoring system according to claim 1, characterized in that: Also includes a prompt module; The prompt module is used to provide prompts to parking lot managers based on the weighted parking probability.

Images