A digital analysis method for warehousing logistics based on cloud computing

Abstract

The present invention relates to the field of data prediction technology, and specifically to a digital analysis method for warehousing and logistics based on cloud computing, including: obtaining storage volume data of each cargo, determining each target curve segment corresponding to the cargo, and determining the stability of the storage volume of the cargo according to the similarity between each target curve segment corresponding to the cargo; analyzing the overall change law of the storage volume data according to the stability of the storage volume and the storage volume data corresponding to the cargo, and determining the first predicted storage volume of the cargo; analyzing the difference between the predicted value and the actual value, determining the prediction accuracy corresponding to the cargo, and determining the prediction weight according to the storage volume data of each cargo and the prediction accuracy; and determining the optimal predicted storage volume in combination with the first predicted storage volume and the prediction weight. The present invention obtains the final storage volume prediction by combining two different prediction methods, which can effectively improve the accuracy of storage volume prediction and is suitable for the field of storage volume data prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of data prediction, and in particular to a digital analysis method for warehousing logistics based on cloud computing. Background Art

[0002] After digital analysis of warehouse logistics, the purchase and shipment strategies of different goods in the warehouse can be adjusted to facilitate intuitive analysis by users. When adjusting the purchase and shipment strategies, the future storage volume situation is predicted based on the changes in the storage volume of goods, and the transfer analysis of goods is performed based on the prediction results.

[0003] The current exponential smoothing algorithm is used to predict storage volume. In the process of implementing the exponential smoothing algorithm, there are artificially defined fixed parameters, such as the smoothing coefficient. If the smoothing coefficient is too large, it means that the model gives too high a weight to new observations, is too sensitive to historical data, and ignores the impact of long-term trends or seasonality, which may cause the forecast results to excessively follow recent fluctuations and the forecast results to be inaccurate; if the smoothing coefficient is too small, it means that the model attaches too much importance to historical data and is insufficiently responsive to new observations, which may cause the model to be unable to capture changes in data trends in a timely manner, causing the forecast to lag behind actual development and reduce the accuracy of the forecast. Summary of the invention

[0004] In order to solve the technical problem that the above-mentioned existing prediction algorithms have low accuracy in predicting storage volume, the purpose of the present invention is to provide a digital analysis method for storage logistics based on cloud computing. The technical solution adopted is as follows:

[0005] An embodiment of the present invention provides a method for digital analysis of warehouse logistics based on cloud computing, the method comprising the following steps:

[0006] Obtain the storage quantity sequence of each cargo in the storage logistics, and divide the storage quantity sequence into each storage quantum sequence, wherein the storage quantum sequence is composed of storage quantity data for each day of the year;

[0007] Take any cargo as the target cargo, determine the target curve segments of each storage quantity subsequence corresponding to the target cargo; determine the storage quantity stability of the target cargo according to the similarity between each target curve segment corresponding to the target cargo;

[0008] Analyze the overall change pattern of storage volume data according to the stability of storage volume and the storage quantity subsequences corresponding to the target goods, and determine the first predicted storage volume of the target goods;

[0009] Analyze the difference between the predicted value and the actual value of the storage quantity according to the storage quantity sequence of the target goods, and determine the prediction accuracy corresponding to the target goods;

[0010] Determine the prediction weight corresponding to the target goods according to the storage quantity sequence of each goods, the prediction accuracy corresponding to the target goods, and the first predicted storage quantity;

[0011] The optimal predicted storage quantity of the target goods is determined according to the second predicted storage quantity obtained from the storage quantity subsequence corresponding to the target goods this year, the predicted weight and the first predicted storage quantity.

[0012] Furthermore, analyzing the difference between the predicted value and the actual value of the storage quantity according to the storage quantity sequence of the target goods to determine the prediction accuracy corresponding to the target goods includes:

[0013] The historical storage quantity data in the storage quantity sequence of the target goods are sequentially combined into historical storage quantity arrays; wherein the historical storage quantity data refers to other storage quantity data in the storage quantity sequence except the storage quantity data of the current day, and the number of data in the historical storage quantity array is greater than a preset number;

[0014] The exponential smoothing algorithm is used to predict each historical storage quantity array to obtain each predicted value; the average value of the difference between each predicted value and its corresponding actual value is calculated, and the average value of the difference is negatively normalized to obtain the prediction accuracy corresponding to the target goods.

[0015] Furthermore, the step of determining the prediction weight corresponding to the target goods according to the storage quantity sequence of each goods, the prediction accuracy corresponding to the target goods, and the first predicted storage quantity includes:

[0016] Determine the storage volume curve corresponding to the storage volume sequence of each cargo, analyze the storage volume difference between the target cargo and the comparison cargo according to the storage volume sequence of each cargo and its storage volume curve, and determine the first correction factor of the prediction accuracy corresponding to the target cargo; the comparison cargo is cargo other than the target cargo;

[0017] Determine a second correction factor of the prediction accuracy corresponding to the target goods according to each prediction value corresponding to the target goods, the actual value of each prediction value, and the first predicted storage quantity;

[0018] The ratio of the first correction factor to the second correction factor is normalized, and the product of the normalized ratio and the prediction accuracy corresponding to the target goods is used as the prediction weight corresponding to the target goods.

[0019] Furthermore, the calculation formula of the first correction factor of the prediction accuracy corresponding to the target goods is:

[0020] Where, QoB ais the first correction factor of the prediction accuracy corresponding to the a-th product, the a-th product is the target product, b is the serial number of the comparison product of the a-th product, B is the number of comparison products of the a-th product, exp is an exponential function with a natural constant as the base, DTW (a,b) is the DTW distance between the storage volume curve of the ath commodity and the storage volume curve of the bth comparison commodity, n is the number of storage volume data in the storage volume sequence, i is the storage volume data sequence number in the storage volume sequence, Ls (a,i) Ls is the i-th storage quantity data in the storage quantity sequence of the a-th cargo, (b,i) is the i-th storage quantity data in the storage quantity sequence of the b-th comparison goods, is the average value of all storage quantity data in the storage quantity sequence of the ath cargo, is the average value of all storage quantity data in the storage quantity sequence of the b-th comparison goods, and || is the absolute value function.

[0021] Furthermore, the calculation formula of the second correction factor of the prediction accuracy corresponding to the target goods is:

[0022] In the formula, Hc a is the second correction factor of the prediction accuracy corresponding to the a-th cargo, the a-th cargo is the target cargo, v is the sequence number of the prediction value, m is the number of prediction values, ΔGz (a,v) is the difference between the vth predicted value corresponding to the ath cargo and the actual value of the vth predicted value, ΔGg (a,t+1) is the difference between the first predicted storage quantity and the second predicted storage quantity of the a-th cargo, t is the current day, t+1 is the day after the current day, and || is the absolute value function;

[0023] Among them, the exponential smoothing algorithm is used to predict the storage quantity sequence corresponding to this year's target goods to obtain the second predicted storage volume.

[0024] Furthermore, the step of determining the target curve segments of each storage subsequence corresponding to the target goods includes:

[0025] Perform curve fitting on each storage quantum sequence corresponding to the target goods to obtain each storage quantum curve; perform clustering processing on each data point on the storage quantum curve to obtain the clustering result; divide the storage quantum curve according to the clustering result to obtain each curve segment of the storage quantum curve; take the sequence number of the day in this year's storage quantum sequence as the target sequence number, and take the curve segment of the storage volume data containing the target sequence number as the target curve segment.

[0026] Furthermore, the determining of the storage quantity stability of the target goods according to the similarity between the target curve segments corresponding to the target goods includes:

[0027] Calculate the similarity between any two target curve segments of the target goods, and use the accumulated value of all similarities as the first storage volume stabilization factor of the target goods;

[0028] Determine the difference between the storage volume data of the target number in the storage sub-sequence of this year and the storage volume data of the target number in the storage sub-sequence of each historical year, accumulate the absolute value of the difference between each difference value and the average value of all the difference values, and perform negative correlation processing on the accumulated values ​​to obtain the second storage volume stability factor of the target goods;

[0029] The product of the first storage quantity stability factor and the second storage quantity stability factor of the target goods is normalized, and the normalized product is used as the storage quantity stability degree of the target goods.

[0030] Further, analyzing the overall change pattern of the storage volume data according to the storage volume stability and each storage volume subsequence corresponding to the target goods to determine the first predicted storage volume of the target goods includes:

[0031] Calculate the absolute value of the difference between two adjacent storage quantity data in the storage quantity subsequence corresponding to the target goods this year, and use the average of all the absolute values ​​of the difference as the adjustment weight for horizontal prediction;

[0032] The next storage quantity data of the storage quantity data of the target sequence number in the storage quantity subsequence of each historical year is selected to form a longitudinal prediction reference sequence, and the average value of the absolute value of the difference between all two adjacent storage quantity data in the longitudinal prediction reference sequence is used as the adjustment weight for longitudinal prediction; wherein, each storage quantity data in the longitudinal prediction reference sequence is arranged in order from far to near in time;

[0033] The first predicted storage volume of the target goods is determined based on the storage volume stability, the storage volume data of the target goods on the day, the adjustment weight during horizontal prediction, the last storage volume data in the vertical prediction reference sequence, and the adjustment weight during vertical prediction.

[0034] Further, the first predicted storage volume of the target goods is determined according to the storage volume stability, the storage volume data of the target goods on the day, the adjustment weight during the horizontal prediction, the last storage volume data in the vertical prediction reference sequence, and the adjustment weight during the vertical prediction, including:

[0035] Pcd (a,t+1) =[Pb (a) *Rv (a,t) *(1+Ge a )]+[(1-Pb (a) )*Ln a *(1+Nca )]; where Pcd (a,t+1)

[0036] is the first predicted storage volume of the a-th cargo, t is the current day, t+1 is the day after the current day, the a-th cargo is the target cargo, Pb (a) is the storage stability of the ath cargo, Rv (a,t) is the storage volume data of the ath cargo on that day, Ge a is the adjustment weight of the ath cargo in horizontal forecasting, Ln a Nc is the last storage quantity data in the longitudinal prediction reference sequence of the ath cargo, a is the adjustment weight of the a-th cargo in the longitudinal forecast.

[0037] Further, determining the optimal predicted storage quantity of the target goods according to the second predicted storage quantity obtained from the storage quantity subsequence corresponding to the target goods this year, the predicted weight and the first predicted storage quantity includes:

[0038] Calculate the product of the second predicted storage volume and the predicted weight as the first product;

[0039] Performing negative correlation processing on the prediction weight, and multiplying the prediction weight after the negative correlation processing and the first predicted storage volume as the second product;

[0040] The value obtained by adding the first product and the second product is used as the optimal predicted storage quantity of the target goods.

[0041] The present invention has the following beneficial effects:

[0042] The present invention provides a digital analysis method for warehousing and logistics based on cloud computing. The method helps to improve the accuracy of storage volume prediction by analyzing the prediction results obtained by two different prediction methods, and further improves the reliability of digital analysis of warehousing and logistics. After obtaining the accuracy of the existing prediction algorithm in predicting the storage volume of target goods through the difference between the predicted value and the actual value, the prediction accuracy is corrected, which can effectively improve the numerical accuracy of the prediction weight corresponding to the target goods, and also provide data support for the subsequent calculation of the optimal storage volume prediction value; when analyzing the overall change law of storage volume data, the stability of the obtained storage volume is also taken into account, which helps to improve the reliability of the first predicted storage volume of the target goods; when calculating the optimal predicted storage volume, the prediction weight is used to balance the prediction situation obtained by the existing prediction algorithm and the first predicted storage volume obtained by the storage volume change law, which effectively enhances the accuracy of the storage volume prediction result and the robustness of the storage volume prediction of different goods. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 This is a flow chart of a digital analysis method for warehousing logistics based on cloud computing according to an embodiment of the present invention;

[0045] Figure 2 The figure is a flowchart of the steps of determining the prediction weight corresponding to the target goods in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0047] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0048] The specific scenarios targeted by the present invention may be:

[0049] By analyzing the correlation between different goods and the changes in storage data, the storage data of goods is predicted. Based on the prediction results, decision makers increase or decrease the storage volume of all goods to ensure the rationality of the flow of goods in the warehouse. The existing exponential smoothing algorithm has artificially limited parameters, which may lead to inaccurate storage volume prediction results, so the prediction results need to be corrected.

[0050] Specifically, this embodiment provides a digital analysis method for warehouse logistics based on cloud computing, such as Figure 1 As shown, the following steps are included:

[0051] S1, obtain the storage quantity sequence of each cargo in the storage logistics, and divide the storage quantity sequence into various storage quantity subsequences.

[0052] The first step is to obtain the storage quantity sequence of each cargo in the warehousing logistics.

[0053] When conducting warehousing and logistics management, it is necessary to collect the storage quantity of each product in the warehouse, which is the inventory situation. Currently, the initial storage quantity sequence of each product can be obtained through optical sensors. The initial storage quantity sequence contains the storage quantity data of each day in 10 years. The frequency and total number of data collection can be set by the implementer according to the actual situation, without specific restrictions.

[0054] In order to reduce abnormal data caused by data collection equipment failure or other special reasons, it is necessary to perform abnormal detection on the initial storage quantity sequence, and smooth the detected abnormal data, and use the initial storage quantity sequence that has completed the smoothing process as the final storage quantity sequence. The final storage quantity sequence can be used for subsequent analysis of storage quantity data changes. The storage quantity sequence of each cargo is the basic data for executing this embodiment. Cargo refers to different types of cargo, such as badminton, badminton rackets, table tennis, football, basketball, etc.

[0055] Among them, the implementation process of anomaly detection can utilize the LOF (Local Outlier Factor) anomaly detection algorithm. The implementation process of the LOF anomaly detection algorithm is a prior art and is not within the protection scope of the present invention, and will not be elaborated here.

[0056] The second step is to divide the storage quantity sequence into individual storage quantum sequences.

[0057] In order to facilitate the subsequent analysis of the change pattern of storage volume of the same goods in different years, it is necessary to divide the storage volume sequence into various storage quantum sequences. The storage quantum sequence is composed of the storage volume data of each day in a year, that is, the storage volume data belonging to the same year in the storage volume sequence are divided together to form a storage quantum sequence. Among them, each storage quantum sequence contains this year's storage quantum sequence.

[0058] It should be noted that after obtaining the storage quantity sequence of each cargo and each storage quantity subsequence, in order to save storage space, all collected data can be uploaded to the cloud data processing system and handed over to the cloud data processing system for subsequent calculations.

[0059] So far, this embodiment has obtained the storage quantity sequence of each cargo in the storage logistics and each storage quantity subsequence.

[0060] S2, taking any cargo as the target cargo, determining the target curve segments of each storage quantity subsequence corresponding to the target cargo; and determining the storage quantity stability of the target cargo according to the similarity between each target curve segment corresponding to the target cargo.

[0061] First of all, it should be explained that the stability of storage volume refers to the stability of the storage volume of the target goods over time. Subsequently, it is necessary to adaptively determine the first predicted storage volume based on the change rules of two adjacent data in the overall storage volume sequence. If the storage volume change stability of the target goods is strong, the importance of the corresponding horizontal prediction parameters when calculating the first predicted storage volume is high. Therefore, the stability of storage volume can characterize the weight in the calculation process of the first predicted storage volume.

[0062] The first step is to determine the target curve segments of each storage quantum sequence corresponding to the target goods.

[0063] Specifically, curve fitting is performed on each storage quantum sequence corresponding to the target goods to obtain each storage quantum curve; clustering is performed on each data point on the storage quantum curve to obtain a clustering result; the storage quantum curve is divided according to the clustering result to obtain each curve segment of the storage quantum curve; the sequence number of the day in this year's storage quantum sequence is used as the target sequence number, and the curve segment containing the storage volume data of the target sequence number is used as the target curve segment.

[0064] It should be noted that in order to longitudinally analyze the similarities between the storage volume change trends of the target goods each year, it is necessary to perform curve fitting on each storage volume sub-sequence; there are special factors affecting the analysis of the storage volume change trend for a whole year. For example, in a certain year, weather factors may cause delivery delays and the storage volume is greatly reduced. In order to reduce the impact of special factors on the similarity analysis between each storage volume sub-curve to a certain extent, clustering processing is performed to locate the subsequent analysis object on the curve segment containing the storage volume data of the target sequence number, that is, the target curve segment.

[0065] Among them, each storage quantum curve has its corresponding target curve segment, and each storage quantum curve must contain the storage quantum curve of the target goods this year. The storage quantum curve closest to the current time is used as this year's storage quantum curve, and the serial number of the day refers to the serial position of today's storage volume data in this year's storage quantum curve. Compared with other storage quantum curve division methods, the clustering method can be used to divide data points with similar storage volume changes into a cluster cluster. The target curve segment obtained in this way can better represent the change characteristics of the storage volume data closest to the current time, which helps to further improve the calculation accuracy of the subsequent storage volume stability, and also provides data conditions for the subsequent storage volume stability calculation.

[0066] Clustering processing can be implemented by DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm, and curve fitting can be implemented by least squares method. The implementation processes of DBSCAN clustering algorithm and least squares method are both prior arts, which are not within the protection scope of the present invention and will not be elaborated in detail this time.

[0067] The second step is to determine the stability of the storage volume of the target goods based on the similarities between the target curve segments corresponding to the target goods.

[0068] In this embodiment, when analyzing the similarity between each target curve segment, the analysis can be conducted from two perspectives: first, the similarity of the overall change trend of any two target curve segments; second, the difference between each historical storage volume data and the storage volume of the day in the same horizontal axis on each target curve segment. The reference value of the storage volume stability obtained from the two perspectives is higher, which is helpful for the subsequent analysis of the change law of the overall storage volume data of the target goods.

[0069] The first sub-step is to calculate the similarity between any two target curve segments of the target goods, and use the accumulated value of all similarities as the first storage volume stabilization factor of the target goods.

[0070] In this embodiment, the similarity between any two target curve segments can be determined by calculating the DTW distance. The smaller the DTW distance, the higher the similarity between the two target curve segments. The first storage volume stability factor of the target goods can be obtained by combining all the DTW distances. The higher the cumulative value of all similarities, the more similar the storage volume changes of the target goods each year, and the larger the first storage volume stability factor.

[0071] It should be noted that the DTW distance is negatively correlated with the similarity, and each DTW distance needs to be negatively correlated, for example, the reciprocal of the DTW distance is used as the similarity between the two target curve segments, or exp(-) is used to achieve it. The calculation process of the DTW distance is a prior art and is not within the scope of protection of the present invention, and will not be elaborated in detail this time.

[0072] The second sub-step is to determine the difference between the storage volume data of the target number in this year's storage volume sub-sequence and the storage volume data of the target number in each historical year's storage volume sub-sequence, accumulate the absolute value of the difference between each difference value and the average value of all difference values, and perform negative correlation processing on the accumulated values ​​to obtain the second storage volume stability factor of the target goods.

[0073] In this embodiment, the storage volume data of the target sequence number is used for vertical difference analysis, which can make the storage volume data of the target sequence number more able to reflect the current storage volume characteristics, which is conducive to improving the numerical accuracy of the second storage volume stability factor. Among them, the target sequence number is the sequence number closest to the predicted sequence number, and the predicted sequence number refers to the next sequence number of the target sequence number in the storage quantum sequence. For this year's storage quantum sequence, the predicted sequence number refers to the sequence number of the predicted data of the next day after the current day. Among them, the historical year is other years except this year.

[0074] As an example, the calculation formula for the second storage volume stability factor of the ath cargo can be:

[0075] Where, Pb2 a is the second storage volume stability factor of the ath cargo, R is the number of historical years, r is the sequence number of the historical year, ΔGe (a,r) is the difference between the storage quantity data of the target number of the a-th cargo in the storage quantum sequence of this year and the storage quantity data of the target number in the storage quantum sequence of the r-th historical year, is the average value of all difference values ​​corresponding to the ath commodity, is the cumulative value of the absolute value of the difference between each difference value corresponding to the ath commodity and the average value of all difference values, For Perform negative correlation processing, | | is the absolute value function.

[0076] In the calculation formula of the second storage volume stability factor, ΔGe (a,r) The smaller it is, the smaller the difference between the storage volume of the ath commodity on that day and the corresponding storage volume in the historical year is, which means that the stability of the target commodity's storage volume in the vertical change is stronger, and the second storage volume stability factor is larger; It can express the difference between the change difference of a single storage volume and the overall change difference. The larger it is, the farther the difference in individual storage volume changes is from the overall change difference, the worse the stability of the difference between the storage volume of the target serial number of all historical years and the storage volume of the day, and the smaller the second storage volume stability factor.

[0077] It should be noted that under normal circumstances There is no zero case. In order to avoid the extreme case where the denominator of the fraction is zero, a hyperparameter can be added to the denominator of the fraction. The hyperparameter can be set to 0.1. The case of adding a hyperparameter to the denominator of the fraction is also applicable to other fractions in this embodiment.

[0078] In the third sub-step, normalize the product of the first storage quantity stability factor and the second storage quantity stability factor of the target goods, and use the normalized product as the storage quantity stability degree of the target goods.

[0079] In this embodiment, firstly, the first storage volume stability factor and the second storage volume stability factor are both positively correlated with the storage volume stability, so the two stability factors can be combined by multiplication; secondly, in order to facilitate the storage volume stability to be used as a weight in subsequent steps, the product of the two stability factors needs to be normalized, and the normalized data is used as the storage volume stability of the target goods. Among them, the implementation process of the normalization process is a prior art and will not be elaborated in detail, and can be implemented by a linear normalization function, that is, a norm function.

[0080] So far, this embodiment has obtained the storage quantity stability of the target goods.

[0081] S3, analyzing the overall change pattern of the storage volume data according to the stability of the storage volume and the storage sub-sequences corresponding to the target goods, and determining the first predicted storage volume of the target goods.

[0082] First, it should be noted that the first predicted storage volume is obtained by analyzing the change rules of the vertical and horizontal storage volume data, and the storage volume stability can be used as a weight to quantify the importance of storage volume data in different directions when calculating the first predicted storage volume. The specific implementation process of determining the first predicted storage volume may include:

[0083] The first step is to calculate the absolute value of the difference between two adjacent storage volume data in the storage volume subsequence corresponding to this year's target goods, and use the average of all the absolute values ​​of the difference as the adjustment weight for horizontal prediction.

[0084] The second step is to select the next storage volume data of the target storage volume data in the storage volume subsequence of each historical year to form a longitudinal prediction reference sequence, and take the average of the absolute values ​​of the differences between all two adjacent storage volume data in the longitudinal prediction reference sequence as the adjustment weight for longitudinal prediction. Among them, the storage volume data in the longitudinal prediction reference sequence are arranged in order from far to near in time.

[0085] The third step is to determine the first predicted storage volume of the target goods based on the stability of the storage volume, the storage volume data of the target goods on the day, the adjustment weight during the horizontal forecast, the last storage volume data in the vertical forecast reference sequence, and the adjustment weight during the vertical forecast.

[0086] As an example, the calculation formula for the first predicted storage quantity of the ath cargo can be:

[0087] Pcd(a,t+1) =[Pb (a) *Rv (a,t) *(1+Ge a )]+[(1-Pb (a) )*Ln a *(1+N ca )]; where Pcd (a,t+1) is the first predicted storage volume of the a-th cargo, t is the current day, t+1 is the day after the current day, the a-th cargo is the target cargo, Pb (a) is the storage stability of the ath cargo, Rv (a,t) is the storage volume data of the ath cargo on that day, Ge a is the adjustment weight of the ath cargo in horizontal forecasting, Ln a Nc is the last storage quantity data in the longitudinal prediction reference sequence of the ath cargo, a is the adjustment weight of the a-th cargo in the longitudinal forecast.

[0088] In the calculation formula of the first predicted storage volume, the adjustment weight for horizontal prediction is obtained by analyzing the storage volume difference of the target goods of this year in two consecutive days, while the adjustment weight for vertical prediction is obtained by analyzing the storage volume difference between the target goods of adjacent historical years in the storage quantum curve of the next sequence number of the target sequence number; Rv (a,t) *(1+Ge a ) can represent the horizontal prediction value obtained based on the change law of storage volume data of this year’s storage quantum sequence, Pb (a) The greater the stability of storage volume, the greater the importance of the horizontal prediction value in calculating the first predicted storage volume; Ln a *(1+Nc a ) can represent the longitudinal prediction value obtained based on the change law of storage volume data of the longitudinal prediction reference sequence of historical years, (1-Pb (a) ) The smaller the negative correlation value of the storage volume stability, the less important the vertical prediction value is when calculating the first predicted storage volume. (a) It can also be used to balance the proportion of horizontal forecast values ​​and vertical forecast values ​​when calculating the first forecast storage volume.

[0089] So far, this embodiment has obtained the initial storage volume of the target goods on the next day after the current day, that is, the first predicted storage volume, by analyzing the change pattern of the overall storage volume.

[0090] S4, analyzing the difference between the predicted value and the actual value of the storage quantity according to the storage quantity sequence of the target goods, and determining the prediction accuracy corresponding to the target goods.

[0091] First of all, it should be noted that the existing prediction algorithm has artificially specified parameters, namely the smoothing coefficient in the exponential smoothing algorithm, which leads to low accuracy in predicting the storage volume of goods. In order to correct the prediction results, it is necessary to understand the accuracy of the prediction results of the storage volume of each cargo when using the existing prediction algorithm, that is, the prediction accuracy corresponding to each cargo.

[0092] In order to facilitate understanding of the implementation process of the entire embodiment, take any one of all the goods in the warehouse as an example, and perform a storage quantity forecast for the next day of the current day, taking any one of the goods as the target goods.

[0093] The first step is to sequentially organize the historical storage quantity data in the storage quantity sequence of the target goods into historical storage quantity arrays.

[0094] In this embodiment, the historical storage quantity data refers to other storage quantity data in the storage quantity sequence except the storage quantity data of the current day. The number of data in the historical storage quantity array is greater than a preset number, and the preset number can be set to 2.

[0095] Preferably, the steps of obtaining each historical storage quantity array may include:

[0096] Assuming that the historical storage quantities in the storage quantity sequence are set to {h1,h2,h3,h4,h5,h6,h7,h8,h9}, the historical storage quantity arrays can be: {h1,h2,h3}, {h1,h2,h3,h4}, {h1,h2,h3,h4,h5}, {h1,h2,h3,h4,h5}, {h1,h2,h3,h4,h5,h6}, {h1,h2,h3,h4,h5,h6,h7}, {h1,h2,h3,h4,h5,h6,h7,h8} and {h1,h2,h3,h4,h5,h6,h7,h8,h9}.

[0097] Of course, implementers may also use other methods to sequentially organize various historical storage quantity data into various historical storage quantity arrays, without specific limitation.

[0098] In the second step, the exponential smoothing algorithm is used to predict each historical storage quantity array to obtain each predicted value; the average value of the difference between each predicted value and its corresponding actual value is calculated, and the average value of the difference is negatively normalized to obtain the prediction accuracy corresponding to the target goods.

[0099] In this embodiment, each historical storage array has its corresponding predicted value and actual value, such as the predicted value of the historical storage array {h1, h2, h3} is h4', and the actual value is h4. The implementation process of the exponential smoothing algorithm is a prior art and is not within the scope of protection of the present invention, and will not be elaborated here.

[0100] As an example, the calculation formula for the prediction accuracy of the ath cargo can be:

[0101] In the formula, Rc a is the prediction accuracy corresponding to the a-th cargo, v is the sequence number of the prediction value, m is the number of prediction values, ΔGz (a,v) is the difference between the v-th predicted value of the a-th cargo and the actual value of the v-th predicted value, and exp is an exponential function with a natural constant as the base.

[0102] In the calculation formula of prediction accuracy, the smaller the difference between the predicted value and the actual value, the more accurate the existing prediction algorithm is in predicting the storage volume of goods, that is, the higher the prediction accuracy, so it is necessary to Negative correlation processing is performed. Negative correlation means that the dependent variable will decrease as the independent variable increases, and the dependent variable will increase as the independent variable decreases. In addition, since the prediction accuracy is used for prediction weight calculation later, it is necessary to After normalization, the prediction accuracy ranges from 0 to 1; exp(-) can be used to Of course, the implementer may also use other methods to achieve the normalization processing of the negative correlation of the data, without specific limitation.

[0103] So far, this embodiment has obtained the prediction accuracy of the target goods.

[0104] S5, determining a prediction weight corresponding to the target goods according to the storage quantity sequence of each goods, the prediction accuracy corresponding to the target goods, and the first predicted storage quantity.

[0105] First, it should be noted that after obtaining the prediction accuracy corresponding to the target goods, in order to improve the data accuracy of the prediction weight, the correlation between the target goods and the comparison goods and the similarity between the data changes of the first predicted storage volume and the historical storage volume are analyzed through the storage volume sequence and prediction of each goods. The flow chart of the steps to determine the prediction weight corresponding to the target goods is as follows: Figure 2 As shown, the following steps are included:

[0106] The first step is to determine the storage volume curve corresponding to the storage volume sequence of each cargo. According to the storage volume sequence of each cargo and its storage volume curve, the storage volume difference between the target cargo and the comparison cargo is analyzed to determine the first correction factor of the prediction accuracy corresponding to the target cargo.

[0107] Among the many goods in the warehouse, there may be certain correlations between different goods, such as badminton and badminton rackets, table tennis balls and table tennis rackets. By analyzing the differences between the storage quantity data of different goods and analyzing the correlation between the target goods and the various comparison goods of the target goods, we can analyze the storage change rules of the target goods from an overall perspective and then make predictions and analyses on the changes in the storage quantity of the goods.

[0108] The first sub-step is to determine the storage quantity curve corresponding to the storage quantity sequence of each commodity.

[0109] In this embodiment, in order to analyze the difference between the overall storage volume data of the target goods and each comparison goods of the target goods, the least square method can be used to perform curve fitting on the storage volume sequence of each goods to obtain a fitting curve for each goods, and the fitting curve is used as the storage volume curve. The implementation process of the least square method is prior art and is not within the scope of protection of the present invention, and will not be elaborated here.

[0110] In the second sub-step, according to the storage quantity sequence and storage quantity curve of each cargo, the storage quantity difference between the target cargo and the comparison cargo is analyzed to determine the first correction factor of the prediction accuracy corresponding to the target cargo.

[0111] In this embodiment, when analyzing the storage volume difference between the target goods and the comparison goods, not only the similarity of the overall data trend change can be considered, but also the stability of the difference data change can be considered. The numerical accuracy of the first correction factor determined by the two aspects is higher, and the first correction factor can characterize the correlation between the target goods and the overall comparison goods. Among them, the comparison goods refer to the goods in the storage other than the target goods.

[0112] As an example, the calculation formula for the first correction factor of the prediction accuracy corresponding to the ath cargo may be:

[0113] Where, QoB a is the first correction factor of the prediction accuracy corresponding to the a-th product, the a-th product is the target product, b is the serial number of the comparison product of the a-th product, B is the number of comparison products of the a-th product, exp is an exponential function with a natural constant as the base, DTW (a,b) is the DTW distance between the storage volume curve of the ath commodity and the storage volume curve of the bth comparison commodity, n is the number of storage volume data in the storage volume sequence, i is the storage volume data sequence number in the storage volume sequence, Ls (a,i) Ls is the i-th storage quantity data in the storage quantity sequence of the a-th cargo, (b,i) is the i-th storage quantity data in the storage quantity sequence of the b-th comparison goods, is the average value of all storage quantity data in the storage quantity sequence of the ath cargo, is the average value of all storage quantity data in the storage quantity sequence of the b-th comparison goods, and || is the absolute value function.

[0114] In the calculation formula of the first correction factor, DTW (a,b) It can represent the overall difference in storage volume between the a-th cargo and the b-th target cargo, DTW (a,b) The smaller it is, the higher the similarity of the storage volume change trends between the two goods is, the higher the correlation is, and the larger the first correction factor is; It can represent the ratio between the ath cargo and the bth comparison cargo on the same day. The smaller it is, the more stable the proportional relationship between the a-th commodity and the b-th comparison commodity is, the higher the correlation will be, and the larger the first correction factor will be; exp(-) can realize the normalization processing of the negative correlation of the data to ensure that the value range of the first correction factor is between 0 and 1.

[0115] In the second step, a second correction factor of the prediction accuracy corresponding to the target goods is determined according to each prediction value corresponding to the target goods, the actual value of each prediction value and the first predicted storage quantity.

[0116] In order to further improve the accuracy of the analysis results of the storage quantity prediction accuracy of the target goods, in this embodiment, by analyzing the difference between the predicted value obtained by the existing prediction algorithm and the corresponding actual value, and the difference between the first predicted storage quantity obtained according to the law of storage quantity data change and the second predicted storage quantity obtained by the existing prediction algorithm, the difference between the first predicted storage quantity and the actual value is quantified. The smaller the difference between the first predicted storage quantity and the actual value, the lower the prediction accuracy of the second predicted storage quantity obtained by the existing prediction algorithm, so there is a negative correlation between the second correction factor and the final prediction weight.

[0117] As an example, the calculation formula for the second correction factor of the prediction accuracy corresponding to the a-th cargo may be:

[0118] In the formula, Hc a is the second correction factor of the prediction accuracy corresponding to the a-th cargo, the a-th cargo is the target cargo, v is the sequence number of the prediction value, m is the number of prediction values, ΔGz (a,v) is the difference between the vth predicted value corresponding to the ath cargo and the actual value of the vth predicted value, ΔGg (a,t+1) is the difference between the first predicted storage quantity and the second predicted storage quantity of the a-th cargo, t is the current day, t+1 is the day after the current day, and || is the absolute value function.

[0119] In the calculation formula of the second correction factor, Indicates the difference between the predicted value obtained by the existing prediction algorithm and its corresponding actual value, ΔGg (a,t+1) Indicates the difference between the first predicted storage volume obtained according to the change law of storage volume data and the second predicted storage volume obtained by the existing prediction algorithm; the second predicted storage volume can be obtained by using the exponential smoothing algorithm to predict the storage volume subsequence corresponding to the target goods this year; The smaller it is, the smaller the difference between the first predicted storage quantity and most actual values ​​is, and the higher the numerical accuracy of the first predicted storage quantity obtained by the storage quantity data change law is. In other words, the lower the prediction accuracy of the second predicted storage quantity obtained by the existing prediction algorithm, and the smaller the second correction factor is.

[0120] In the third step, the ratio of the first correction factor to the second correction factor is normalized, and the product of the normalized ratio and the prediction accuracy corresponding to the target goods is used as the prediction weight corresponding to the target goods.

[0121] The first correction factor is positively correlated with the prediction weight. The higher the first correction factor is, the greater the impact of the storage volume of the comparison goods on the prediction accuracy of the target goods, and the greater the degree to which the prediction accuracy corresponding to the target goods is adjusted upward; the second correction factor is negatively correlated with the prediction weight. The greater the difference between the first predicted storage volume and the majority of actual values, the greater the degree to which the prediction accuracy corresponding to the target goods is adjusted downward in order to adjust the accuracy of the prediction results of the existing prediction algorithm of the target goods.

[0122] In this embodiment, the ratio of the first correction factor to the second correction factor is normalized again in order to ensure the numerical regularity of the final prediction weight, so as to facilitate the subsequent importance analysis of the prediction values ​​obtained by different calculation methods; the normalized ratio is used as the coefficient of the prediction accuracy corresponding to the target goods, which can improve the calculation accuracy of the prediction weight.

[0123] Among them, the existing normalization implementation process is a prior art, which is not within the protection scope of the present invention and will not be elaborated in detail. The normalization implementation method can be a linear normalization function.

[0124] So far, this embodiment has obtained the prediction weight when predicting the storage quantity of the target goods on the next day after the current day.

[0125] S6, determining the optimal predicted storage quantity of the target goods according to the second predicted storage quantity, the predicted weight and the first predicted storage quantity obtained from the storage quantity subsequence corresponding to the target goods this year.

[0126] It should be noted that after obtaining the first predicted storage volume, in order to further improve the accuracy of the prediction result of the target goods, the second predicted storage volume obtained by the existing prediction algorithm is combined to obtain the optimal predicted storage volume with higher accuracy. The specific implementation process may include:

[0127] The first step is to calculate the product of the second predicted storage volume and the predicted weight as the first product.

[0128] In the second step, the prediction weight is negatively correlated, and the product of the prediction weight after negative correlation processing and the first predicted storage volume is used as the second product.

[0129] The third step is to use the value obtained by adding the first product and the second product as the optimal predicted storage quantity of the target goods.

[0130] It should be noted that the prediction weight is obtained by analyzing the existing prediction algorithm, which can represent the prediction accuracy of the existing prediction algorithm. The higher the prediction accuracy of the existing prediction algorithm, the more important the second predicted storage volume obtained by the exponential smoothing algorithm is in calculating the optimal predicted storage volume; otherwise, the more important the first predicted storage volume obtained by the comprehensive historical storage volume data change law is. The accuracy of the optimal predicted storage volume obtained by the two prediction algorithms is higher, which is more conducive to the subsequent accurate adjustment of the purchase and shipment of the target goods based on the prediction results.

[0131] This concludes the present embodiment.

[0132] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A digital analysis method for warehousing logistics based on cloud computing, characterized in that: The following steps are involved: Obtain the storage quantity sequence of each cargo in the storage logistics, and divide the storage quantity sequence into each storage quantum sequence, wherein the storage quantum sequence is composed of storage quantity data for each day of the year; Take any cargo as the target cargo, determine the target curve segments of each storage quantity subsequence corresponding to the target cargo; determine the storage quantity stability of the target cargo according to the similarity between each target curve segment corresponding to the target cargo; Analyze the overall change pattern of storage volume data according to the stability of storage volume and the storage quantity subsequences corresponding to the target goods, and determine the first predicted storage volume of the target goods; Analyze the difference between the predicted value and the actual value of the storage quantity according to the storage quantity sequence of the target goods, and determine the prediction accuracy corresponding to the target goods; Determine the prediction weight corresponding to the target goods according to the storage quantity sequence of each goods, the prediction accuracy corresponding to the target goods, and the first predicted storage quantity; Determine the optimal predicted storage volume of the target goods according to the second predicted storage volume obtained from the storage subsequence corresponding to the target goods this year, the predicted weight and the first predicted storage volume; The step of determining the target curve segments of each storage quantum sequence corresponding to the target goods includes: Perform curve fitting on each storage quantum sequence corresponding to the target goods to obtain each storage quantum curve; perform clustering processing on each data point on the storage quantum curve to obtain a clustering result; divide the storage quantum curve according to the clustering result to obtain each curve segment of the storage quantum curve; take the sequence number of the day in the storage quantum sequence of this year as the target sequence number, and take the curve segment of the storage volume data containing the target sequence number as the target curve segment; Analyzing the difference between the predicted value and the actual value of the storage quantity according to the storage quantity sequence of the target goods to determine the prediction accuracy corresponding to the target goods includes: The historical storage quantity data in the storage quantity sequence of the target goods are sequentially combined into historical storage quantity arrays; wherein the historical storage quantity data refers to other storage quantity data in the storage quantity sequence except the storage quantity data of the current day, and the number of data in the historical storage quantity array is greater than a preset number; Use the exponential smoothing algorithm to predict each historical storage quantity array to obtain each predicted value; calculate the average value of the difference between each predicted value and its corresponding actual value, perform negative correlation normalization processing on the average value of the difference, and obtain the prediction accuracy corresponding to the target goods; The step of determining the prediction weight corresponding to the target goods according to the storage quantity sequence of each goods, the prediction accuracy corresponding to the target goods, and the first predicted storage quantity includes: Determine the storage volume curve corresponding to the storage volume sequence of each cargo, analyze the storage volume difference between the target cargo and the comparison cargo according to the storage volume sequence of each cargo and its storage volume curve, and determine the first correction factor of the prediction accuracy corresponding to the target cargo; the comparison cargo is cargo other than the target cargo; Determine a second correction factor of the prediction accuracy corresponding to the target goods according to each prediction value corresponding to the target goods, the actual value of each prediction value, and the first predicted storage quantity; Normalizing the ratio of the first correction factor to the second correction factor, and taking the product of the normalized ratio and the prediction accuracy corresponding to the target goods as the prediction weight corresponding to the target goods; The calculation formula of the first correction factor of the prediction accuracy corresponding to the target goods is: Where, QoB a is the first correction factor of the prediction accuracy corresponding to the a-th product, the a-th product is the target product, b is the serial number of the comparison product of the a-th product, B is the number of comparison products of the a-th product, exp is an exponential function with a natural constant as the base, DTW (a,b) is the DTW distance between the storage volume curve of the ath commodity and the storage volume curve of the bth comparison commodity, n is the number of storage volume data in the storage volume sequence, i is the storage volume data sequence number in the storage volume sequence, Ls (a,i) Ls is the i-th storage quantity data in the storage quantity sequence of the a-th cargo, (b,i) is the i-th storage quantity data in the storage quantity sequence of the b-th comparison goods, is the average value of all storage quantity data in the storage quantity sequence of the ath cargo, is the average value of all storage quantity data in the storage quantity sequence of the b-th comparison goods, | | is the absolute value function; The calculation formula of the second correction factor of the prediction accuracy corresponding to the target goods is: In the formula, Hc a is the second correction factor of the prediction accuracy corresponding to the a-th cargo, the a-th cargo is the target cargo, v is the sequence number of the prediction value, m is the number of prediction values, ΔGz (a,v) is the difference between the vth predicted value corresponding to the ath cargo and the actual value of the vth predicted value, ΔGg (a,t+1) is the difference between the first predicted storage quantity and the second predicted storage quantity of the a-th cargo, t is the current day, t+1 is the day after the current day, | | is the absolute value function; Among them, the exponential smoothing algorithm is used to predict the storage quantity sequence corresponding to the target goods this year to obtain the second predicted storage quantity; The step of analyzing the overall change pattern of the storage volume data according to the storage volume stability and each storage volume subsequence corresponding to the target goods to determine the first predicted storage volume of the target goods includes: Calculate the absolute value of the difference between two adjacent storage quantity data in the storage quantity subsequence corresponding to the target goods this year, and use the average of all the absolute values ​​of the difference as the adjustment weight for horizontal prediction; The next storage quantity data of the storage quantity data of the target sequence number in the storage quantity subsequence of each historical year is selected to form a longitudinal prediction reference sequence, and the average value of the absolute value of the difference between all two adjacent storage quantity data in the longitudinal prediction reference sequence is used as the adjustment weight for longitudinal prediction; wherein, each storage quantity data in the longitudinal prediction reference sequence is arranged in order from far to near in time; Determine the first predicted storage volume of the target goods according to the storage volume stability, the storage volume data of the target goods on the day, the adjustment weight during the horizontal forecast, the last storage volume data in the vertical forecast reference sequence, and the adjustment weight during the vertical forecast; The method of determining the first predicted storage volume of the target goods according to the storage volume stability, the storage volume data of the target goods on the day, the adjustment weight during the horizontal prediction, the last storage volume data in the vertical prediction reference sequence, and the adjustment weight during the vertical prediction includes: Pcd (a,t+1) =[Pb (a) *Rv (a,t) *(1+Ge a )]+[(1-Pb (a) )*Ln a *(1+Nc a )]; where Pcd (a,t+1) is the first predicted storage volume of the a-th cargo, t is the current day, t+1 is the day after the current day, the a-th cargo is the target cargo, Pb (a) is the storage stability of the ath cargo, Rv (a,t) is the storage volume data of the ath cargo on that day, Ge a is the adjustment weight of the ath cargo in horizontal forecasting, Ln a Nc is the last storage quantity data in the longitudinal prediction reference sequence of the ath cargo, a is the adjustment weight of the ath cargo in the longitudinal forecast; The determining the optimal predicted storage quantity of the target goods according to the second predicted storage quantity obtained from the storage quantity subsequence corresponding to the target goods this year, the predicted weight and the first predicted storage quantity comprises: Calculate the product of the second predicted storage volume and the predicted weight as the first product; Performing negative correlation processing on the prediction weight, and multiplying the prediction weight after the negative correlation processing and the first predicted storage volume as the second product; The value obtained by adding the first product and the second product is used as the optimal predicted storage quantity of the target goods.

2. According to the cloud computing-based warehousing logistics digital analysis method of claim 1, it is characterized in that: The step of determining the storage volume stability of the target goods according to the similarity between the target curve segments corresponding to the target goods includes: Calculate the similarity between any two target curve segments of the target goods, and use the accumulated value of all similarities as the first storage volume stabilization factor of the target goods; Determine the difference between the storage volume data of the target number in the storage sub-sequence of this year and the storage volume data of the target number in the storage sub-sequence of each historical year, accumulate the absolute value of the difference between each difference value and the average value of all the difference values, and perform negative correlation processing on the accumulated values ​​to obtain the second storage volume stability factor of the target goods; The product of the first storage quantity stability factor and the second storage quantity stability factor of the target goods is normalized, and the normalized product is used as the storage quantity stability degree of the target goods.

Images