CN115511408A - Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof - Google Patents

Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof Download PDF

Info

Publication number
CN115511408A
CN115511408A CN202211144120.XA CN202211144120A CN115511408A CN 115511408 A CN115511408 A CN 115511408A CN 202211144120 A CN202211144120 A CN 202211144120A CN 115511408 A CN115511408 A CN 115511408A
Authority
CN
China
Prior art keywords
medicine
price
early warning
index
model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202211144120.XA
Other languages
Chinese (zh)
Inventor
魏新江
蔡文鼎
曲帅
张慧凤
邵喜高
迟颖
曹晓宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ludong University
Original Assignee
Ludong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ludong University filed Critical Ludong University
Priority to CN202211144120.XA priority Critical patent/CN115511408A/en
Publication of CN115511408A publication Critical patent/CN115511408A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
    • G06Q10/087—Inventory or stock management, e.g. order filling, procurement or balancing against orders
    • G06Q10/0875—Itemisation or classification of parts, supplies or services, e.g. bill of materials
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data
    • G06Q30/0206—Price or cost determination based on market factors
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Theoretical Computer Science (AREA)
  • General Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Marketing (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Artificial Intelligence (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Primary Health Care (AREA)
  • Tourism & Hospitality (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Quality & Reliability (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

A medicine centralized purchasing monitoring and early warning visual platform and a monitoring and early warning method thereof belong to the technical field of medicine management. The problems of inaccurate medicine collection, monitoring and early warning and low precision are solved. The invention collects the time sequence data of the order form of the centralized medicine purchase, the annual section data of the centralized medicine purchase; the medicine price monitoring method comprises the following steps: establishing a Laplace price index, a Pair price index and a chain Laplace price index by using the sorted data, and then analyzing the fluctuation rule of the medicine price; predicting the price index through an ARIMA model and a Holt two-parameter index smoothing model; carrying out regression prediction on the purchase price of the medicine through an LASSO regression model based on the order quantity and the arrival quantity data; the medicine shortage early warning method comprises the following steps: establishing a medicine shortage early warning index by using the processed data; and establishing a medicine shortage classification prediction model based on machine learning, statistical theory and deep learning. The invention ensures the prediction precision and the prediction stability.

Description

Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof
Technical Field
The invention belongs to the technical field of medicine management, and particularly relates to a medicine centralized purchasing monitoring and early warning visual platform and a monitoring and early warning method thereof.
Background
The Centralized purchasing (Centralized purchasing) monitoring and early warning of the medicines is a work with strong specialty and technicality, and is a work with high requirements on programs, methods and the like. The market price monitoring department closely pays attention to market dynamics through monitoring and early warning, analyzes and predicts the price trend of the medicine, achieves sensitive and accurate information and timely and powerful measures, and well plays a management guiding role. Therefore, the construction of the monitoring and early warning system for the centralized medicine purchase has great significance for perfecting a centralized medicine bid-attracting purchase mechanism and a price forming mechanism taking the market as a leading factor, stabilizing the price of the medicine market, meeting the medicine purchase demand and practically reducing the medicine burden of the masses. However, the centralized medicine purchasing data has the characteristics of large scale, complex structure and the like, and if an improper processing mode is adopted for the complex, various and large-scale data, huge cost and great energy are often spent, and the obtained data analysis conclusion is difficult to achieve an ideal effect. In order to overcome the problems, the medicine collection monitoring and early warning mainly adopts qualitative analysis, policy exploration, description statistics, literature research methods and the like; in the aspects of medicine price trend prediction and medicine shortage early warning, common methods such as factor analysis and multiple linear regression analysis are mostly adopted. For monitoring, describing and counting the centralized purchasing process of the medicine, a simple machine learning method cannot find the problems of medicine price fluctuation and abnormal medicine supply in time, methods such as policy exploration, literature research and the like have poor timeliness, and can not make instant prediction and feedback on abnormal conditions in the current or future purchasing process.
Disclosure of Invention
The invention aims to solve the problems of inaccurate medicine collection monitoring and early warning and low precision, and provides a medicine centralized purchasing monitoring and early warning visual platform and a monitoring and early warning method thereof.
A monitoring and early warning method of a medicine centralized purchasing monitoring and early warning visual platform comprises the following steps:
s1, collecting time sequence data of a concentrated medicine purchase order and concentrated medicine purchase annual section data;
s2, monitoring the price of the medicine:
s2.1, performing data sorting on the time sequence data of the concentrated medicine purchase order collected in the step S1;
s2.2, establishing a Laplace price index, a Party price index and a chained Laplace price index by using the sorted data, and then analyzing the fluctuation rule of the medicine price;
s2.3, forecasting the price index through an ARIMA model and a Holt double-parameter index smoothing model;
s2.4, carrying out regression prediction on the purchase price of the medicine through an LASSO regression model based on the order quantity and the arrival quantity data;
s3, medicine shortage early warning:
s3.1, carrying out data processing on the section data of the concentrated medicine purchasing year collected in the step S1;
s3.2, establishing a medicine shortage early warning index by using the processed data;
and S3.3, establishing a medicine shortage classification prediction model based on machine learning, statistical theory and deep learning.
Further, the data arrangement in step S2.1 includes extracting the total amount of the monthly order of the drug, the arrival quantity, the arrival rate, the drug purchase transaction price and the number of the price-reduced items, wherein the drug purchase price extracting method includes the following steps:
s2.1.1 matching the medicine to be researched with the medicine centralized purchasing platform medicine, and screening medicine information;
s2.1.2, when the common name is used as a sample, unifying the drug metering units, converting the drug dosage and the drug price into the drug frequency DDDs and the limited daily cost DDDc by using the limited daily dose DDD, and calculating the drug price change amplitude and the drug saving amount, wherein the calculation formula is as follows:
DDDs = total amount of drug used over a certain period of time/DDD value of drug
DDDs Change Width = ((Current time DDDs-last time DDDs)/last time DDDs) × 100%
DDDc = total purchase of drug/drug DDDs
DDDc change width = ((current period DDDc-last period DDDc)/last period DDDc) × 100%
Medicine saved amount = current period DDDs × (last period DDDc-current period DDDc);
s2.1.3, when the product code is used as the research object, matching is performed in the platform screen-hung drug ID field every month without conversion.
Further, in step S2.2, the larch price index, the pascal price index, and the chain larch price index are established by using the total monthly order amount and the drug purchase transaction price data:
s2.2.1, pascal price index:
Figure BDA0003854596860000021
wherein L is p Representing the Party price index, Q, of the drug purchased in the present period t Indicating the number of sales of the drug at the present date, P 0 Indicating the price of the drug purchased at the base date, P t The price of the medicine purchased in the current period is shown, n is the number of purchase periods based on the base period, and t is the number of the current period;
s2.2.2, larch price index:
Figure BDA0003854596860000031
wherein L is L Indicating the Laplace price index, Q, of the currently purchased drug 0 Indicating the sales quantity of the drugs in the base period;
Figure BDA0003854596860000032
wherein L is F Representing the Fischer price index of the currently purchased medicine;
s2.2.3, chain lat index:
Figure BDA0003854596860000033
wherein L is t ,L t-1 The chain Laplace price index, W, of the purchased medicine in the current stage and the previous stage respectively t-1 Weight, P, representing the previous period t-1 Representing the price of the medicine purchased in the previous period;
s2.2.4, analyzing the fluctuation rule of the drug price: the Party price index is the number of the medicines in the current period as the weight compared with the previous period and is used for the change of the expense cost caused by the current period change of the consumption structure; the Laplace price index takes the quantity of the base period as the weight and reflects the change of the pure price of the medicine; the Fischer price index averages different bias of pull type and Party; the chain Laplace price index reflects the structural changes of all medicines, including the changes of the price and the quantity of the medicines; and selecting the chain Laplace price index and the Fisher price index as models used for final price index prediction.
Further, the specific method for predicting the price index through the ARIMA model and the Holt two-parameter index smoothing model in the step S2.3 comprises the following steps:
s2.3.1, stationarity test of ARIMA model:
judging whether the price index original sequence is stable or not, adopting an ADF (automatic document surface) inspection method, inspecting that the price index original sequence is not stable, and performing first-order difference to ensure that the data is stable;
s2.3.2 and ARIMA model construction: and (3) determining the order of the model by observing an autocorrelation coefficient and a partial autocorrelation coefficient graph, wherein the ARIMA model structural expression is as follows:
Figure BDA0003854596860000034
wherein phi (B) = 1-phi 1 B-…-φ p B p ,Θ(B)＝1-θ 1 B-…-θ q B q Are the autoregressive coefficients and moving average coefficient polynomials of ARMA (p, q), respectively, B is the delay operator, phi i Is coefficient weight, i =1, …, p, θ j Is the coefficient weight, j =1, …, q; u is the mean of the sequence after the difference, ε t Is a zero mean white noise sequence;
Figure BDA0003854596860000041
representing the sequence d-order difference, x, by the delay operator B t Representing an original sequence of price indices;
s2.3.3, model parameter test: establishing a model as an ARIMA (1,1,2) model by observing the autocorrelation coefficient and the partial autocorrelation coefficient graph, and establishing the model for parameter inspection;
s2.3.4, residual sequence test: the Q statistic was chosen for pure randomness test and is expressed as follows:
Figure BDA0003854596860000042
wherein c is the sequence observation period number; m is a designated delay period number, rho is a sample autocorrelation coefficient, and when the Q statistic is greater than a 1-alpha quantile point of chi-square distribution with the degree of freedom m or the P value of the statistic is less than alpha, the sequence is a non-white noise sequence;
s2.3.5, using S2.3.1-S2.3.4 to predict price index;
s2.3.6 and Holt two-parameter exponential smoothing model construction: the model structure is represented as follows:
Figure BDA0003854596860000043
wherein:
Figure BDA0003854596860000044
and
Figure BDA0003854596860000045
the intercept term and the slope term are continuously smoothed according to the latest observed value of the sequence; x is a radical of a fluorine atom t The latest observed value of the sequence at the time t; alpha and beta are smooth coefficients and have a value range of [0,1]；
Prediction of e epochs
Figure BDA0003854596860000046
Comprises the following steps:
Figure BDA0003854596860000047
the results of the model establishment in this example, the α and β results are 1 and 0.7507, respectively, and the prediction expression of the Holt two-parameter exponential smoothing model obtained is as follows:
Figure BDA0003854596860000048
further, in step S2.4, the concrete implementation method of performing regression prediction on the purchase price of the medicine through the LASSO regression model based on the order quantity and the arrival quantity data is as follows:
s2.4.1, the objective function L (b) of the LASSO regression model is:
L(b)＝∑(y-Xb) 2 +λ||b|| 1 ＝∑(y-Xb) 2 +∑λ|b|
wherein λ | | b | | non-luminous 1 Is a penalty term of a function, lambda is a penalty coefficient, | | b | | luminance 1 The index is the regular of a regression coefficient b and represents the sum of absolute values of all regression coefficients, a dependent variable y is a chain Laplace price index of the chronic disease drugs sold on a platform in a network every month, and an independent variable X is the arrival rate, the price reduction type number, the national statistical bureau and other authoritative index data;
s2.4.2, obtaining the optimal lambda value 0.0051 by a ten-fold cross validation method, solving the regression coefficient of the model to obtain the CPI index with penalty of 0, removing the CPI independent variable, and completing the processScreening to obtain model prediction index data for regression prediction analysis of medicine purchase price, and using R 2 And evaluating the model prediction result by using the MAE evaluation index.
Further, the method for processing the data of the section data of the concentrated procurement year of the medicine collected in the step S1 in the step S3.1 comprises the following steps: the method comprises the steps of sorting and visualizing related information of the medicine shortage, drawing sales volume, arrival rate and supply and demand curves, collecting concentrated purchasing data of the medicine of nearly five years for sorting, and extracting average response time of arrival rate, sales volume, unit price, delivery of business enterprises and receiving of medical institutions and medicine attributes of the medicine of three days and five days as early warning indexes for subsequently establishing a medicine shortage classification prediction model.
Further, the method for establishing the drug shortage warning index by using the processed data in step S3.2 includes the following steps:
s3.2.1, performing K-S test on the early warning index obtained in the step S3.1 to obtain a conclusion that the P value of the early warning index is far less than the significance level of 0.05, which indicates that the early warning indexes obtained in the step S3.1 are not in accordance with normal distribution;
using Mann-Whitney U for inspection to obtain indexes with significant difference between the shortage medicines and the non-shortage medicines, wherein except the arrival rate, other indexes are inspected, so that the arrival rate index is eliminated, and early warning indexes except the arrival rate index are used for model construction;
s3.2.2, analyzing and extracting common factors by using the factors, synthesizing the early warning indexes into common indexes, and establishing a medicine shortage early warning model:
firstly, KMO and Bartlett sphericity test are used to judge whether factor analysis is applicable to a sample, KMO statistic is 0.566, P value of the Bartlett sphericity test is 0.000 which is far less than significance level 0.05, and factor analysis is indicated to be applicable to the sample;
secondly, factor rotation is carried out by adopting a maximum variance orthogonal rotation method, the original multiple early warning indexes are integrated into 5 public factors, and a public factor linear expression of 15 standardized indexes is obtained according to a public factor score coefficient:
F 1 ＝0.054X 1 +0.046X 2 +0.003X 3 +0.005X 4 -0.027X 5 +0.057X 6 +0.227X 7 -0.176X 8 +0.041X 9 -0.158X 10 -0.123X 11 +0.161X 12 +0.252X 13 +0.227X 14 +0.351X 15 ，
F 2 ＝0.193X 1 -0.018X 2 -0.105X 3 +0.048X 4 +0.118X 5 +0.019X 6 +0.069X 7 +0.034X 8 -0.011X 9 +0.428X 10 +0.433X 11 -0.015X 12 +0.050X 13 -0.175X 14 -0.165X 15 ，
F 3 ＝-0.274X 1 +0.069X 2 -0.023X 3 +0.098X 4 +0.197X 5 +0.218X 6 -0.112X 7 +0.521X 8 +0.305X 9 +0.129X 10 +0.047X 11 +0.035X 12 -0.162X 13 +0.000X 14 -0.219X 15 ，
F 4 ＝-0.111X 1 +0.170X 2 +0.695X 3 -0.358X 4 +0.288X 5 -0.035X 6 +0.001X 7 -0.045X 8 +0.006X 9 -0.055X 10 -0.086X 11 -0.026X 12 +0.005X 13 -0.034X 14 +0.011X 15 ，
F 5 ＝-0.211X 1 +0.682X 2 +0.124X 3 +0.469X 4 +0.090X 5 -0.076X 6 +0.099X 7 +0.014X 8 +0.148X 9 -0.004X 10 +0.055X 11 -0.100X 12 +0.078X 13 -0.051X 14 +0.020X 15 。
wherein the linear expression calculates a common factor F i (i =1,2, …, 5), X represents an index.
Further, the method for establishing the medicine shortage classification prediction model based on machine learning, statistical theory and deep learning in step S3.3 includes the following steps:
s3.3.1, construction P is a summary of the risk of medicine shortageThe rate, the value range is 0-1, and the factors influencing the shortage value of the medicine are set and recorded as x 1 ,x 2 ,...,x k To do so by
Figure BDA0003854596860000061
For dependent variables, a linear regression equation was established as follows:
Figure BDA0003854596860000062
the mathematical expression of the Logistic model is as follows:
Figure BDA0003854596860000063
s3.3.2 substituting linear expression to calculate common factor to obtain F 2 And F 3 P value of (A) is greater than significance level 0.05, indicating that F 2 And F 3 The effect is not significant, so the effect is removed to finally obtain the F 1 ，F 4 And F 5 The Logistic medicine shortage classification prediction model is as follows:
Figure BDA0003854596860000064
s3.3.3, respectively substituting the training sample and the test sample data into Logistic medicine shortage classification prediction model in S3.3.2 for verification, wherein the division point is set to 0.5, if the probability P corresponding to the tested medicine is greater than 0.5, the medicine is determined as a medicine in shortage, otherwise, the medicine is not a medicine in shortage.
A medicine centralized purchasing monitoring and early warning visualization platform comprises a medicine price monitoring module and a medicine shortage early warning module, wherein the medicine price monitoring module comprises a price index module, a current situation analysis module and a price fluctuation early warning module, and the medicine price monitoring module comprises a current situation analysis module, a goods arrival rate module and a medicine shortage prediction module;
the drug price monitoring module is used for compiling Laplace, paris and chain Laplace price indexes of different types, analyzing the fluctuation rule of the current centralized purchased drug price, predicting the price indexes through an ARIMA time sequence model and a neural network algorithm, and performing regression prediction on the drug purchase price through machine learning algorithms such as LASSO and a support vector regression machine by using the order quantity and the arrival quantity data;
the medicine shortage early warning module is used for analyzing the current situation of medicine shortage according to medicine centralized purchasing data, researching main influence factors of the medicine shortage, designing a three-day arrival rate and average response time medicine shortage early warning index, and predicting whether the medicine is in shortage or not through Logistic regression and a random forest machine learning algorithm.
The invention has the beneficial effects that:
the invention relates to a monitoring and early warning method of a medicine centralized purchasing monitoring and early warning visual platform, which comprises the following steps:
based on the multidimensional and disordered big data of medicine centralized purchase, preprocessing work such as data cleaning and index extraction is completed, the influence factors of medicine price fluctuation and medicine supply abnormity are discussed by referring to the experiences of relevant experts and medicine enterprise management personnel, and indexes reflecting the medicine price level and the medicine shortage condition are designed.
Designing a price index model based on the time sequence data of the concentrated purchase order of the medicine and by utilizing information such as the consignment price, the order quantity, the arrival quantity and the like of the medicine, extracting indexes influencing the price fluctuation of the medicine, and establishing a medicine price fluctuation trend regression prediction model based on machine learning, statistical theory and deep learning;
based on the section data of the concentrated medicine purchasing year, the indexes influencing medicine supply are extracted by utilizing information such as medicine transaction, medicine attribute and the like, and a medicine shortage classification prediction model based on machine learning, statistical theory and deep learning is established.
The invention provides a monitoring and early warning system based on machine learning, statistical theory and deep learning, which comprises the following components:
1) A medicine price monitoring system is constructed, different types of price indexes such as Laplace, paris, chain Laplace and the like are compiled, the fluctuation rule of the medicine price is analyzed, and the price indexes are predicted through an ARIMA time sequence model and a neural network algorithm. And then, by using information such as order quantity, arrival quantity and the like, carrying out regression prediction on the purchase price of the medicine through machine learning algorithms such as LASSO, a support vector regression machine and the like. Based on the analysis, the fluctuation trend of the centralized purchase price of the medicine is accurately predicted, and reasonable suggestions are provided for governments and enterprises to seek for stabilizing the price of the medicine.
2) And constructing a medicine shortage early warning system, analyzing the current situation of medicine shortage according to the related data of medicine centralized purchase, and exploring main influence factors of the medicine shortage. And (3) designing a medicine shortage early warning index by combining policy documents and related expert guidance, and predicting whether the medicine is in shortage or not through a Logiti regression machine learning algorithm. Based on the analysis, the medicine shortage condition of centralized medicine purchase is accurately predicted, and reasonable suggestions are provided for governments and enterprises to guarantee medicine supply.
The invention provides a visualization platform for monitoring and early warning of centralized medicine purchasing, which comprises:
the system at least comprises a medicine price monitoring module and a medicine shortage early warning module, and is used for deeply mining medicine centralized purchasing data, wherein the medicine price monitoring module comprises medicine price index compiling and predicting, medicine price fluctuation regression predicting and the like, the medicine shortage early warning module comprises medicine shortage current situation analysis, medicine shortage prediction analysis and the like, and a visual webpage platform for medicine centralized purchasing monitoring and early warning is established. The system provides decision basis and powerful technical support for public resource transaction centers and related departments of various provinces, provides price reference for purchasing parties, production parties and governments, and monitors the price change of the medicines and the abnormal conditions of medicine supply in real time.
The invention provides a monitoring and early warning method, a system and a visual platform based on machine learning, statistical theory and deep learning. By sorting the related data of concentrated medicine purchase and combining the experience of related experts and medicine enterprise managers, the influence factors of medicine price fluctuation and medicine supply abnormity are discussed, and indexes reflecting medicine price level and medicine shortage condition are designed. Based on the time sequence data and the annual section data of the concentrated purchase order of the medicine, by utilizing information such as the price of the finished product of the medicine, the quantity of the order, the quantity of the arrived goods, the attribute of the medicine and the like, a price index model is designed, indexes influencing the price fluctuation and the medicine supply of the medicine are extracted, and a regression prediction model of the price fluctuation trend of the medicine and a classification prediction model of the medicine shortage based on machine learning, statistical theory and deep learning are established. The problems of price fluctuation and abnormal supply of the current concentrated medicine can be effectively fed back in time while the prediction precision and the prediction stability are ensured.
Drawings
Fig. 1 is a flowchart of a monitoring and early warning method of a medicine centralized procurement monitoring and early warning visualization platform according to the present invention;
fig. 2 is a schematic structural diagram of a centralized drug procurement monitoring and early warning visualization platform according to the present invention;
FIG. 3 is a Party price index chart of the monitoring and early warning method of the visualization platform for centralized drug procurement monitoring and early warning according to the present invention;
FIG. 4 is a chain Laplace price index diagram of the monitoring and early warning method of the visualization platform for centralized drug procurement monitoring and early warning according to the present invention;
fig. 5 is a relation diagram of regular coefficients and regression coefficients in a LASSO regression model of the monitoring and early warning method of the visualization platform for centralized drug procurement monitoring and early warning according to the present invention.
FIG. 6 is a comparison graph of the LASSO regression model prediction results of the monitoring and early warning method of the visualization platform for centralized drug procurement monitoring and early warning according to the present invention;
FIG. 7 is a drug price monitoring flow chart of a monitoring and early warning method of a drug centralized procurement monitoring and early warning visualization platform according to the invention;
fig. 8 is a medicine shortage early warning flow chart of the monitoring early warning method of the medicine centralized procurement monitoring early warning visualization platform according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and the detailed description. It is to be understood that the embodiments described herein are illustrative only and are not limiting, i.e., that the embodiments described are only a few embodiments, rather than all, of the present invention. While the components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations, the present invention is capable of other embodiments.
Thus, the following detailed description of specific embodiments of the present invention, presented in the accompanying drawings, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the detailed description of the invention without inventive step, are within the scope of protection of the invention.
For a further understanding of the contents, features and effects of the present invention, the following embodiments will be illustrated in detail with reference to the accompanying drawings 1-8:
the first embodiment is as follows:
a monitoring and early warning method for a medicine centralized purchasing monitoring and early warning visual platform comprises the following steps:
s1, collecting time sequence data of a medicine centralized purchase order and annual section data of the medicine centralized purchase;
s2, monitoring the price of the medicine:
s2.1, performing data sorting on the time sequence data of the concentrated purchase order of the medicine collected in the step S1;
further, the data arrangement in step S2.1 includes extracting the total amount of the monthly order of the drug, the arrival quantity, the arrival rate, the drug purchase transaction price and the number of the price-reduced items, wherein the drug purchase price extracting method includes the following steps:
s2.1.1 matching the medicine to be researched with the medicine centralized purchasing platform medicine, and screening medicine information;
s2.1.2, when the common name is used as a sample, unifying the drug metering units, converting the drug dosage and the drug price into the drug frequency DDDs and the limited daily cost DDDc by using the limited daily dose DDD, and calculating the drug price change amplitude and the drug saving amount, wherein the calculation formula is as follows:
DDDs = total amount of drug used over a certain period of time/DDD value of drug
DDDs Change Width = ((Current time DDDs-last time DDDs)/last time DDDs) × 100%
DDDc = total purchase of drug/drug DDDs
DDDc change width = ((current period DDDc-last period DDDc)/last period DDDc) × 100%
Medicine saved amount = current period DDDs × (last period DDDc-current period DDDc);
DDDs refer to the number of days that a person can purchase the medicine, and reflect the frequency of use of a certain medicine. The larger the value is, the higher the use frequency of the medicine is, and the medicine is a clinical common variety;
DDDc is an economic parameter of the medicine, represents the total price level of the medicine and represents the average daily cost for taking the medicine;
DDDs and DDDc are used for replacing the dosage and price of the medicine, the medicine has the advantages of being free from the influence of different sale prices, packages, specifications and daily dosages of various medicines, and solving the problem that the price cannot be compared due to different dosages of different medicines for one time and different times of daily administration.
For example:
the DDD value of the olanzapine tablet is 10mg, the total amount of the olanzapine tablet used in 2 months of 2020 is 17804820, the total amount of the olanzapine tablet purchased is 6197073.98, and then the olanzapine tablet is prepared
DDDs＝17804820/10＝1780482，DDDc＝6197073.98/1780482＝3.480559747。
Such as: calculating the change amplitude of DDDs and DDDc of the olanzapine tablet by taking years as units:
the olanzapine tablet has DDDs of 180.4582, DDDc of 15.2474 in the year before the belt amount procurement is carried out, DDDs of 248.8508, DDDc of 4.0545 in the year after the belt amount procurement is carried out,
then DDDs change amplitude = ((248.8508-180.4582)/180.4582) × 100% =37.90%,
DDDc variation width = ((4.0545-15.2474)/15.2474) × 100% = -73.41%.
The result shows that the frequency of administration of olanzapine tablets is increased by 37.90% and the limited daily cost is reduced by 73.41% after the administration of the olanzapine tablets is purchased.
Such as: calculating the saving amount of olanzapine tablets by taking the year as a unit:
the saved amount (unit: yuan) =248.8508 x (15.2474-4.0545) =2785.362119, which indicates that the olanzapine tablet is saved by 2785.362119 yuan after the tape amount purchase is carried out compared with before the tape amount purchase is carried out.
S2.1.3, when the product code is used as the research object, matching is carried out on the ID field of the drug sold on the platform in a net hanging manner according to each month, and conversion is not carried out;
s2.2, establishing a Laplace price index, a Party price index and a chained Laplace price index by using the sorted data, and then analyzing the fluctuation rule of the medicine price;
further, in step S2.2, the larch price index, the pascal price index, and the chain larch price index are established by using the total monthly order amount and the drug purchase transaction price data:
s2.2.1, pascal price index:
Figure BDA0003854596860000101
wherein L is p Representing the Party price index, Q, of the drug purchased in the present period t Indicating the number of sales of the drug at the present date, P 0 Indicating the price of the drug purchased at the base date, P t The price of the medicine purchased in the current period is represented, n is the period number of purchase based on the base period, and t represents the current period number;
s2.2.2, larch price index:
Figure BDA0003854596860000102
wherein L is L Expressing the Laplace price index, Q, of the currently purchased medicine 0 Indicating the sales quantity of the drugs in the base period;
Figure BDA0003854596860000111
wherein L is F Representing the Fischer price index of the currently purchased medicine;
s2.2.3, chain lat index:
Figure BDA0003854596860000112
wherein L is t ,L t-1 The chain Laplace price index, W, of the purchased medicine in the current stage and the previous stage respectively t-1 Weight, P, representing the previous period t-1 Representing the price of the medicine purchased in the previous period;
s2.2.4, analyzing the fluctuation rule of the drug price: the Party price index is the number of the medicines in the current period as the weight compared with the previous period and is used for the change of the expense cost caused by the current period change of the consumption structure; the Laplace price index takes the quantity of the base period as the weight and reflects the change of the pure price of the medicine; the Fischer price index averages different bias of pull type and Party; the chain Laplace price index reflects the structural changes of all medicines, including the changes of the price and the quantity of the medicines; selecting a chain Laplace price index and a Fisher price index as models used for predicting the final price index;
s2.3, predicting the price index through an ARIMA model and a Holt two-parameter index smoothing model;
further, the specific method for predicting the price index through the ARIMA model and the Holt two-parameter index smoothing model in the step S2.3 comprises the following steps:
s2.3.1, stationarity test of ARIMA model:
judging whether the price index original sequence is stable or not, adopting an ADF (automatic document surface) inspection method, inspecting that the price index original sequence is not stable, and performing first-order difference to ensure that the data is stable;
s2.3.2 and ARIMA model construction: and (3) determining the model by observing the autocorrelation coefficient and the partial autocorrelation coefficient graph, wherein the ARIMA model structure expression is as follows:
Figure BDA0003854596860000113
wherein phi (B) = 1-phi 1 B-…-φ p B p ,Θ(B)＝1-θ 1 B-…-θ q B q Are the autoregressive coefficients and moving average coefficient polynomials of ARMA (p, q), respectively, B is the delay operator, phi i Is coefficient weight, i =1, …, p, θ j Is the coefficient weight, j =1, …, q; u is the mean of the sequence after the difference, ε t A zero mean white noise sequence;
Figure BDA0003854596860000114
representing the d-order difference, x, of the sequence by the delay operator B t Representing an original sequence of price indices;
s2.3.3, model parameter test: establishing a model as an ARIMA (1,1,2) model by observing the autocorrelation coefficient and the partial autocorrelation coefficient graph, and establishing the model for parameter inspection;
s2.3.4, residual sequence test: q statistic is selected to carry out a pure randomness test, and the expression of the Q statistic is as follows:
Figure BDA0003854596860000121
wherein n is the sequence observation period number (based on the purchase period number of the base period, if one meaning, the purchase period number is deleted, if not one meaning, the letter needs to be replaced); m is a designated delay period number, rho is a sample autocorrelation coefficient, and when the Q statistic is greater than a 1-alpha quantile point of chi-square distribution with the degree of freedom m or the P value of the statistic is less than alpha, the sequence is a non-white noise sequence;
s2.3.5, using S2.3.1-S2.3.4 to predict price index;
further, the construction and prediction process is described in detail as follows:
firstly, observing whether an original sequence is stable or not, wherein the sequence is not stable, performing first-order differential correction, and checking stability; then observing an autocorrelation coefficient graph and a partial autocorrelation coefficient graph according to the sequence after the first-order difference, determining the order of the model, and initially constructing the model; then, according to the established model, performing model inspection and parameter t inspection on the model, wherein the model is meaningful when the established ARIMA (1,1,2) passes the inspection; finally, carrying out pure randomness test on the residual sequence, and fully extracting sequence information after passing the test; and predicting the price index sequence according to the established model, presenting a steady and descending trend, and explaining the effectiveness of the concentrated medicine purchasing policy implementation.
S2.3.6 and Holt two-parameter exponential smoothing model construction: the model structure is represented as follows:
Figure BDA0003854596860000122
wherein:
Figure BDA0003854596860000123
and
Figure BDA0003854596860000124
the intercept term and the slope term are continuously smoothed according to the latest observed value of the sequence; x is the number of t The latest observed value of the sequence at the time t; alpha and beta are smooth coefficients and have a value range of [0,1]；
Prediction of e epochs
Figure BDA0003854596860000125
Comprises the following steps:
Figure BDA0003854596860000126
the results of the model establishment of this example, alpha and beta, are 1 and 0.7507, respectively, and the prediction expression of the Holt two-parameter exponential smoothing model is obtained as follows:
Figure BDA0003854596860000127
s2.4, carrying out regression prediction on the purchase price of the medicine through an LASSO regression model based on the order quantity and the arrival quantity data;
further, in step S2.4, the concrete implementation method of performing regression prediction on the purchase price of the medicine through the LASSO regression model based on the order quantity and the arrival quantity data is as follows:
s2.4.1, the objective function L (b) of the LASSO regression model is:
L(b)＝∑(y-Xb) 2 +λ||b|| 1 ＝∑(y-Xb) 2 +∑λ|b|
wherein λ | | b | | non-luminous 1 Is a penalty term of a function, lambda is a penalty coefficient, | | b | | luminance 1 The index is the regular expression of a regression coefficient b and represents the sum of absolute values of all regression coefficients, a dependent variable y is a chain Laplace price index of the chronic disease drugs sold on a platform in a network every month, and an independent variable X is one of the arrival rate, the price reduction type number and the relevant index data of the State statistics bureau;
s2.4.2, obtaining the optimal lambda value of 0.0051 by a ten-fold cross validation method, solving the regression coefficient of the model to obtain the CPI index which is punished to 0, eliminating the CPI independent variable, completing screening to obtain model prediction index data, performing regression prediction analysis on the purchase price of the medicine, and performing R cross validation on the model prediction index data to obtain the optimal lambda value of 0.0051 2 And evaluating the model prediction result by using the MAE evaluation index;
as can be seen from FIG. 5, the regular and regression coefficients appear as a trumpet fold line, indicating that a variable multicollinearity condition exists.
The variables screened with LASSO as new independent variable sets were predicted with other models, based on which the results of prediction with the ANN model are shown in fig. 6: as can be seen from FIG. 6, the model predicts a result R 2 And the MAE evaluation indexes are 97% and 0.6 respectively, and compared with a general regression prediction model, the data obtained in the same way has better prediction effect.
S3, medicine shortage early warning:
s3.1, carrying out data processing on the section data of the concentrated medicine purchasing year collected in the step S1;
further, the method for processing the data of the section data of the concentrated procurement years of the medicines collected in the step S1 in the step S3.1 comprises the following steps: arranging and visualizing related information of the shortage medicines, drawing sales volume, arrival rate and supply and demand curves, collecting concentrated purchasing data of medicines of nearly five years for arrangement, and extracting average response time of arrival rate, sales volume, unit price, delivery of business enterprises and receiving of medical institutions and medicine attributes of three days and five days as early warning indexes for subsequently establishing a medicine shortage classification prediction model;
s3.2, establishing a medicine shortage early warning index by using the processed data;
further, the method for establishing the drug shortage warning index by using the processed data in step S3.2 includes the following steps:
s3.2.1, performing K-S test on the early warning index obtained in the step S3.1 to obtain a conclusion that the P value of the early warning index is far less than the significance level 0.05, which indicates that the early warning index obtained in the step S3.1 does not conform to normal distribution;
using Mann-WhitneyU for inspection to obtain indexes with significant difference between the shortage medicines and the non-shortage medicines, and excluding the arrival rate, other indexes are inspected, so that the arrival rate index is eliminated, and using early warning indexes except the arrival rate index for model construction;
s3.2.2, analyzing and extracting common factors by using the factors, synthesizing the early warning indexes into common indexes, and establishing a medicine shortage early warning model:
firstly, using KMO and Bartlett spherical test to judge whether factor analysis is suitable for a sample, wherein the KMO statistic is 0.566, the P value of the Bartlett spherical test is 0.000 and is far less than the significance level 0.05, and the factor analysis is suitable for the sample;
secondly, factor rotation is carried out by adopting a maximum variance orthogonal rotation method, the original multiple early warning indexes are integrated into 5 public factors, and a public factor linear expression of 15 standardized indexes is obtained according to a public factor score coefficient:
F 1 ＝0.054X 1 +0.046X 2 +0.003X 3 +0.005X 4 -0.027X 5 +0.057X 6 +0.227X 7 -0.176X 8 +0.041X 9 -0.158X 10 -0.123X 11 +0.161X 12 +0.252X 13 +0.227X 14 +0.351X 15 ，
F 2 ＝0.193X 1 -0.018X 2 -0.105X 3 +0.048X 4 +0.118X 5 +0.019X 6 +0.069X 7 +0.034X 8 -0.011X 9 +0.428X 10 +0.433X 11 -0.015X 12 +0.050X 13 -0.175X 14 -0.165X 15 ，
F 3 ＝-0.274X 1 +0.069X 2 -0.023X 3 +0.098X 4 +0.197X 5 +0.218X 6 -0.112X 7 +0.521X 8 +0.305X 9 +0.129X 10 +0.047X 11 +0.035X 12 -0.162X 13 +0.000X 14 -0.219X 15 ，
F 4 ＝-0.111X 1 +0.170X 2 +0.695X 3 -0.358X 4 +0.288X 5 -0.035X 6 +0.001X 7 -0.045X 8 +0.006X 9 -0.055X 10 -0.086X 11 -0.026X 12 +0.005X 13 -0.034X 14 +0.011X 15 ，
F 5 ＝-0.211X 1 +0.682X 2 +0.124X 3 +0.469X 4 +0.090X 5 -0.076X 6 +0.099X 7 +0.014X 8 +0.148X 9 -0.004X 10 +0.055X 11 -0.100X 12 +0.078X 13 -0.051X 14 +0.020X 15 。
wherein the linear expression calculates a common factor F i (i =1,2, …, 5), X represents an index;
s3.3, establishing a medicine shortage classification prediction model based on machine learning, statistical theory and deep learning;
further, the method for establishing the medicine shortage classification prediction model based on machine learning, statistical theory and deep learning in step S3.3 includes the following steps:
s3.3.1, P is constructed as the probability of medicine shortage risk, the value range is 0-1, and the factors influencing the medicine shortage value are set and recorded as x 1 ,x 2 ,...,x k To do so by
Figure BDA0003854596860000141
For dependent variables, a linear regression equation was established as follows:
Figure BDA0003854596860000142
the mathematical expression of the Logistic model is as follows:
Figure BDA0003854596860000143
s3.3.2, substituting linear expression to calculate common factor to obtain F 2 And F 3 P value of (A) is greater than significance level 0.05, indicating that F 2 And F 3 The effect is not significant, so the effect is removed to finally obtain the F 1 ，F 4 And F 5 The Logistic medicine shortage classification prediction model is as follows:
Figure BDA0003854596860000151
s3.3.3, respectively substituting the training sample and the test sample data into Logistic medicine shortage classification prediction model in S3.3.2 for verification, wherein the division point is set to 0.5, if the probability P corresponding to the tested medicine is greater than 0.5, the medicine is determined as a medicine in shortage, otherwise, the medicine is not a medicine in shortage.
Furthermore, the prediction accuracy of the obtained training sample and the test sample data is 97.06% and 91.18% respectively. From the prediction precision of the model in the training sample and the test sample, the model has good fitting effect on historical data, and can also accurately predict future data, which shows that the established early warning model for the medicine shortage has good prediction performance.
The second embodiment is as follows:
a medicine centralized purchasing monitoring and early warning visual platform comprises a medicine price monitoring module and a shortage medicine early warning module, wherein the medicine price monitoring module comprises a price index module, a current situation analysis module and a price fluctuation early warning module, and the medicine price monitoring module comprises a current situation analysis module, a delivery rate module and a medicine shortage prediction module;
the drug price monitoring module is used for compiling Laplace, paris and chain Laplace price indexes of different types, analyzing the fluctuation rule of the current centralized purchased drug price, predicting the price indexes through an ARIMA time sequence model and a neural network algorithm, and performing regression prediction on the drug purchase price through machine learning algorithms such as LASSO and a support vector regression machine by using the order quantity and the arrival quantity data;
the medicine shortage early warning module is used for analyzing the current medicine shortage situation according to the centralized medicine purchasing data, exploring main influence factors of medicine shortage, designing the three-day arrival rate and average response time medicine shortage early warning indexes, and predicting whether medicines are in shortage or not through Logistic regression and a random forest machine learning algorithm.
Furthermore, the visualized platform for monitoring and early warning of centralized medicine purchasing is provided with a medicine price monitoring module and a medicine shortage early warning module, and is used for deeply mining the centralized medicine purchasing data, wherein the medicine price monitoring module comprises medicine price index compiling and predicting, medicine price fluctuation regression predicting and the like, the medicine shortage early warning module comprises medicine shortage current situation analyzing, medicine shortage predicting analyzing and the like, and the visualized webpage platform for monitoring and early warning of centralized medicine purchasing is established. The system provides decision basis and powerful technical support for public resource trading centers and related departments, provides price reference for buyers, producers and governments, and monitors the price change of medicines and the abnormal condition of medicine supply in real time.
It is noted that relational terms such as "first" and "second," and the like, may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising a … …" does not exclude the presence of another identical element in a process, method, article, or apparatus that comprises the element.
While the application has been described above with reference to specific embodiments, various modifications may be made and equivalents may be substituted for elements thereof without departing from the scope of the application. In particular, the various features of the embodiments disclosed herein may be used in any combination that is not inconsistent with the structure, and the failure to exhaustively describe such combinations in this specification is merely for brevity and resource conservation. Therefore, it is intended that the application not be limited to the particular embodiments disclosed, but that the application will include all embodiments falling within the scope of the appended claims.

Claims (9)

1. A monitoring and early warning method of a medicine centralized purchasing monitoring and early warning visual platform is characterized by comprising the following steps: the method comprises the following steps:
s1, collecting time sequence data of a medicine centralized purchase order and annual section data of the medicine centralized purchase;
s2, monitoring the price of the medicine:
s2.1, performing data sorting on the time sequence data of the concentrated purchase order of the medicine collected in the step S1;
s2.2, establishing a Laplace price index, a Paris price index and a chain Laplace price index by using the sorted data, and then analyzing the fluctuation rule of the medicine price;
s2.3, predicting the price index through an ARIMA model and a Holt two-parameter index smoothing model;
s2.4, carrying out regression prediction on the purchase price of the medicine through an LASSO regression model based on the order quantity and the arrival quantity data;
s3, medicine shortage early warning:
s3.1, carrying out data processing on the section data of the concentrated medicine purchasing year collected in the step S1;
s3.2, establishing a medicine shortage early warning index by using the processed data;
and S3.3, establishing a medicine shortage classification prediction model based on machine learning, statistical theory and deep learning.
2. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform of claim 1, wherein: the data arrangement in the step S2.1 comprises the steps of extracting the total amount of the monthly order of the medicine, the arrival quantity, the arrival rate, the purchase transaction price of the medicine and the number of the price-reduced products, wherein the medicine purchase price extracting method comprises the following steps:
s2.1.1 matching the medicine to be researched with the medicine centralized purchasing platform medicine, and screening medicine information;
s2.1.2, when the common name is used as a sample, unifying the drug metering units, converting the drug dosage and the drug price into the drug frequency DDDs and the limited daily cost DDDc by using the limited daily dose DDD, and calculating the drug price change amplitude and the drug saving amount, wherein the calculation formula is as follows:
DDDs = total amount of usage over a certain period of time of drug/DDD value of drug
DDDs Change Width = ((Current time DDDs-last time DDDs)/last time DDDs) × 100%
DDDc = total purchase of drug/drug DDDs
DDDc change width = ((current period DDDc-last period DDDc)/last period DDDc) × 100%
Medicine saved amount = current period DDDs × (last period DDDc-current period DDDc);
s2.1.3, when the product code is used as the research object, matching is performed in the platform screen-hung drug ID field every month without conversion.
3. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform as claimed in claim 1 or 2, wherein: in step S2.2, the total amount of the orders per month and the drug purchase transaction price data are utilized to establish a Laplace price index, a Party price index and a chained Laplace price index:
s2.2.1, pascal price index:
Figure FDA0003854596850000021
wherein L is p Showing the Paris price index, Q, of the drug purchased at the present time t Indicating the number of sales of the drug at the present date, P 0 Indicating the price of the drug purchased at the base date, P t The price of the medicine purchased in the current period is shown, n is the number of purchase periods based on the base period, and t is the number of the current period;
s2.2.2, larch price index:
Figure FDA0003854596850000022
wherein L is L Indicating the Laplace price index, Q, of the currently purchased drug 0 Indicating the sales quantity of the drugs in the base period;
Figure FDA0003854596850000023
wherein L is F Representing the Fischer price index of the currently purchased medicine;
s2.2.3, chain lat price index:
Figure FDA0003854596850000024
wherein L is t ,L t-1 The chain Laplace price index, W, of the purchased medicine in the current stage and the previous stage respectively t-1 Weight, P, representing the previous period t-1 Representing the price of the medicine purchased in the previous period;
s2.2.4, analyzing the fluctuation rule of the drug price: the Party price index is the weight of the current drug quantity as compared with the previous period and is used for the change of expense cost caused by the current change of the consumption structure; the Laplace price index takes the quantity of the base period as the weight and reflects the change of the pure price of the medicine; the Fischer price index averages different bias of pull type and Party; the chain Laplace price index reflects the structural changes of all medicines, including the changes of the price and the quantity of the medicines; and selecting the chain Laplace price index and the Fisher price index as models used for final price index prediction.
4. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform of claim 3, wherein: step S2.3 the concrete method for predicting the price index through the ARIMA model and the Holt double-parameter index smoothing model comprises the following steps:
s2.3.1, stationarity test of ARIMA model:
judging whether the price index original sequence is stable or not, adopting an ADF (automatic document surface) inspection method, inspecting that the price index original sequence is not stable, and performing first-order difference to ensure that the data is stable;
s2.3.2 and ARIMA model construction: and (3) determining the order of the model by observing an autocorrelation coefficient and a partial autocorrelation coefficient graph, wherein the ARIMA model structural expression is as follows:
Figure FDA0003854596850000031
wherein phi (B) = 1-phi 1 B-…-φ p B p ,Θ(B)＝1-θ 1 B-…-θ q B q Are the autoregressive coefficients and moving average coefficient polynomials of ARMA (p, q), respectively, B is the delay operator, phi i Is coefficient weight, i =1, …, p, θ j Is the coefficient weight, j =1, …, q; u is the mean of the sequence after differentiation, ε t Is a zero mean white noise sequence;
Figure FDA0003854596850000032
representing the d-order difference, x, of the sequence by the delay operator B t Representing an original sequence of price indices;
s2.3.3, model parameter test: establishing a model as an ARIMA (1,1,2) model by observing the autocorrelation coefficient and the partial autocorrelation coefficient graph, and establishing the model for parameter inspection;
s2.3.4, residual sequence test: the Q statistic was chosen for pure randomness test and is expressed as follows:
Figure FDA0003854596850000033
wherein c is the sequence observation period number, m is the designated delay period number, rho is the sample autocorrelation coefficient, and when the Q statistic is greater than the 1-alpha quantile point of chi-square distribution with the degree of freedom m or the P value of the statistic is less than alpha, the sequence is a non-white noise sequence;
s2.3.5, using S2.3.1-S2.3.4 to predict price index;
s2.3.6 and Holt two-parameter exponential smoothing model construction: the model structure is represented as follows:
Figure FDA0003854596850000034
wherein:
Figure FDA0003854596850000035
and
Figure FDA0003854596850000036
the intercept term and the slope term are continuously smoothed according to the latest observed value of the sequence; x is the number of t The latest observed value of the sequence at the time t; alpha and beta are smooth coefficients and have a value range of [0,1]；
Prediction of e epochs
Figure FDA0003854596850000037
Comprises the following steps:
Figure FDA0003854596850000038
the results of the model establishment in this example, the α and β results are 1 and 0.7507, respectively, and the prediction expression of the Holt two-parameter exponential smoothing model obtained is as follows:
Figure FDA0003854596850000041
5. the monitoring and early warning method of the medicine centralized procurement monitoring and early warning visualization platform of claim 4, characterized in that: s2.4 the concrete implementation method for carrying out regression prediction on the purchase price of the medicine through the LASSO regression model based on the order quantity and the arrival quantity data comprises the following steps:
s2.4.1, the objective function L (b) of the LASSO regression model is:
L(b)＝∑(y-Xb) 2 +λ||b|| 1 ＝∑(y-Xb) 2 +∑λ|b|
wherein λ | | b | | non-luminous 1 Is a penalty term of the function, lambda is a penalty coefficient, | b | | calness 1 The system is a regular form of a regression coefficient b and represents the sum of absolute values of all regression coefficients, a dependent variable y is a chain Laplace price index of the chronic disease drugs sold on a platform in a network every month, and an independent variable X is the arrival rate, the price reduction type number, the national statistical bureau and other authoritative index data;
s2.4.2, obtaining the optimal lambda value of 0.0051 by a ten-fold cross validation method, solving the regression coefficient of the model to obtain the CPI index which is punished to 0, eliminating the CPI independent variable, completing screening to obtain model prediction index data, performing regression prediction analysis on the purchase price of the medicine, and performing R cross validation on the model prediction index data to obtain the optimal lambda value of 0.0051 2 And evaluating the model prediction result by using the MAE evaluation index.
6. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform of claim 5, wherein: step S3.1 the method for data processing of the data of the section of the concentrated procurement year of the medicine collected in step S1 is as follows: the method comprises the steps of sorting and visualizing shortage medicine information, drawing sales volume, arrival rate and supply and demand curves, collecting concentrated medicine purchasing data of nearly five years for sorting, and extracting average response time and medicine attributes of arrival rate, sales volume, unit price, business delivery and medical institution receiving of three days and five days as early warning indexes for subsequently establishing a medicine shortage classification prediction model.
7. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform of claim 6, wherein: step S3.2 the method for establishing the medicine shortage early warning index by utilizing the processed data comprises the following steps:
s3.2.1, performing K-S test on the early warning index obtained in the step S3.1 to obtain a conclusion that the P value of the early warning index is far less than the significance level 0.05, which indicates that the early warning index obtained in the step S3.1 does not conform to normal distribution;
using Mann-WhitneyU for inspection to obtain indexes with significant difference between the shortage medicines and the non-shortage medicines, and excluding the arrival rate, other indexes are inspected, so that the arrival rate index is eliminated, and using early warning indexes except the arrival rate index for model construction;
s3.2.2, analyzing and extracting common factors by using the factors, synthesizing the early warning indexes into common indexes, and establishing a medicine shortage early warning model:
firstly, using KMO and Bartlett spherical test to judge whether factor analysis is suitable for a sample, wherein the KMO statistic is 0.566, the P value of the Bartlett spherical test is 0.000 and is far less than the significance level 0.05, and the factor analysis is suitable for the sample;
secondly, factor rotation is carried out by adopting a maximum variance orthogonal rotation method, the original multiple early warning indexes are integrated into 5 public factors, and a public factor linear expression of 15 standardized indexes is obtained according to a public factor score coefficient:
F 1 ＝0.054X 1 +0.046X 2 +0.003X 3 +0.005X 4 -0.027X 5 +0.057X 6 +0.227X 7 -0.176X 8 +0.041X 9 -0.158X 10 -0.123X 11 +0.161X 12 +0.252X 13 +0.227X 14 +0.351X 15 ，
F 2 ＝0.193X 1 -0.018X 2 -0.105X 3 +0.048X 4 +0.118X 5 +0.019X 6 +0.069X 7 +0.034X 8 -0.011X 9 +0.428X 10 +0.433X 11 -0.015X 12 +0.050X 13 -0.175X 14 -0.165X 15 ，
F 3 ＝-0.274X 1 +0.069X 2 -0.023X 3 +0.098X 4 +0.197X 5 +0.218X 6 -0.112X 7 +0.521X 8 +0.305X 9 +0.129X 10 +0.047X 11 +0.035X 12 -0.162X 13 +0.000X 14 -0.219X 15 ，
F 4 ＝-0.111X 1 +0.170X 2 +0.695X 3 -0.358X 4 +0.288X 5 -0.035X 6 +0.001X 7 -0.045X 8 +0.006X 9 -0.055X 10 -0.086X 11 -0.026X 12 +0.005X 13 -0.034X 14 +0.011X 15 ，
F 5 ＝-0.211X 1 +0.682X 2 +0.124X 3 +0.469X 4 +0.090X 5 -0.076X 6 +0.099X 7 +0.014X 8 +0.148X 9 -0.004X 10 +0.055X 11 -0.100X 12 +0.078X 13 -0.051X 14 +0.020X 15 。
wherein the linear expression calculates a common factor F i (i =1,2, …, 5), X represents an index.
8. The monitoring and early warning method of the centralized medicine procurement monitoring and early warning visualization platform of claim 7, wherein: step S3.3 the method for establishing the medicine shortage classification prediction model based on machine learning, statistical theory and deep learning comprises the following steps:
s3.3.1, P is the probability of medicine shortage risk, the value range is 0-1, and the factors influencing the medicine shortage value are set and recorded as x 1 ,x 2 ,...,x k To do so by
Figure FDA0003854596850000051
For dependent variables, a linear regression equation was established as follows:
Figure FDA0003854596850000052
the mathematical expression of the Logistic model is as follows:
Figure FDA0003854596850000053
s3.3.2, substituting linear expression to calculate common factor to obtain F 2 And F 3 P value of (A) is greater than significance level 0.05, indicating F 2 And F 3 The effect is not significant, so the effect is removed to finally obtain the F 1 ，F 4 And F 5 The Logistic medicine shortage classification prediction model is as follows:
Figure FDA0003854596850000054
s3.3.3, respectively substituting the training sample and the test sample data into Logistic medicine shortage classification prediction model in S3.3.2 for verification, wherein the division point is set to 0.5, if the probability P corresponding to the tested medicine is greater than 0.5, the medicine is determined as a medicine in shortage, otherwise, the medicine is not a medicine in shortage.
9. The utility model provides a purchase monitoring early warning visual platform is concentrated to medicine which characterized in that: the system comprises a medicine price monitoring module and a medicine shortage early warning module, wherein the medicine price monitoring module comprises a price index module, a current situation analysis module and a price fluctuation early warning module, and the medicine price monitoring module comprises a current situation analysis module, a delivery rate module and a medicine shortage prediction module;
the drug price monitoring module is used for compiling Laplace, paris and chain Laplace price indexes of different types, analyzing the fluctuation rule of the current centralized purchased drug price, predicting the price indexes through an ARIMA time sequence model and a neural network algorithm, and performing regression prediction on the drug purchase price through machine learning algorithms such as LASSO and a support vector regression machine by using the order quantity and the arrival quantity data;
the medicine shortage early warning module is used for analyzing the current situation of medicine shortage according to medicine centralized purchasing data, researching main influence factors of the medicine shortage, designing a three-day arrival rate and average response time medicine shortage early warning index, and predicting whether the medicine is in shortage or not through Logistic regression and a random forest machine learning algorithm.
CN202211144120.XA 2022-09-20 2022-09-20 Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof Pending CN115511408A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211144120.XA CN115511408A (en) 2022-09-20 2022-09-20 Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211144120.XA CN115511408A (en) 2022-09-20 2022-09-20 Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof

Publications (1)

Publication Number Publication Date
CN115511408A true CN115511408A (en) 2022-12-23

Family

ID=84503272

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211144120.XA Pending CN115511408A (en) 2022-09-20 2022-09-20 Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof

Country Status (1)

Country Link
CN (1) CN115511408A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116342171A (en) * 2023-05-31 2023-06-27 北京中科江南信息技术股份有限公司 Method, device and equipment for monitoring supply price of multistage component of communication product

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116342171A (en) * 2023-05-31 2023-06-27 北京中科江南信息技术股份有限公司 Method, device and equipment for monitoring supply price of multistage component of communication product
CN116342171B (en) * 2023-05-31 2023-08-29 北京中科江南信息技术股份有限公司 Method, device and equipment for monitoring supply price of multistage component of communication product

Similar Documents

Publication Publication Date Title
Moges et al. A multidimensional analysis of data quality for credit risk management: New insights and challenges
US9489633B2 (en) Trust rating metric for future event prediction of an outcome
Ehsani-Moghaddam et al. Data quality in healthcare: A report of practical experience with the Canadian Primary Care Sentinel Surveillance Network data
CN107851295B (en) Data analysis apparatus, data analysis method, and storage medium storing data analysis program
CN108182963A (en) A kind of medical data processing method and processing device
JPH07302202A (en) System and method for determination of fuzzy-logic entity operating profile
CN115511408A (en) Medicine centralized purchasing monitoring and early warning visual platform and monitoring and early warning method thereof
CN113362116B (en) Medicine market scale prediction system based on machine learning
Ouyang et al. Estimating intra-cluster correlation coefficients for planning longitudinal cluster randomized trials: a tutorial
Fu et al. Resilient supply chain framework for semiconductor distribution and an empirical study of demand risk inference
Gaebert et al. Efficiency development in the German pharmaceutical market
Gillan et al. Expediting assessments of database performance for streams of respiratory parameters
CN117114812A (en) Financial product recommendation method and device for enterprises
Sahai et al. Pocket dictionary of statistics
Aryandani et al. Implementation of Fuzzy C-Means in Investor Group in the Stock Market Post-Covid-19 Pandemic
CN113642669B (en) Feature analysis-based fraud prevention detection method, device, equipment and storage medium
CN113919891A (en) Medical service project standard cost accounting system and method based on time-estimation operation cost method
CN112785443A (en) Financial product pushing method and device based on client group
CN114595945A (en) Manufacturing process-oriented health degree evaluation method for MBOM (moving object model)
Hochkamp et al. Outlier detection in data mining: Exclusion of errors or loss of information?
Ivanoti et al. Decision Support System for Predicting Employee Leave Using the Light Gradient Boosting Machine (Lightgbm) and K-Means Algorithm
Dieni et al. Business Intelligence for Unemployment Rate Management System
Zhirnova et al. Estimation Of Costs For Metrological Support Of Production
Saarenmaa Assessing and measuring data quality in credit risk modelling
Pang et al. Application of Big Data Information Platform in Medical Equipment.

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination