CN105976199A - Medicine sales prediction method and medicine sales prediction system based on hybrid model - Google Patents
Medicine sales prediction method and medicine sales prediction system based on hybrid model Download PDFInfo
- Publication number
- CN105976199A CN105976199A CN201610264506.2A CN201610264506A CN105976199A CN 105976199 A CN105976199 A CN 105976199A CN 201610264506 A CN201610264506 A CN 201610264506A CN 105976199 A CN105976199 A CN 105976199A
- Authority
- CN
- China
- Prior art keywords
- sales volume
- arima
- data sequence
- model
- historical data
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- 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
- G06Q30/0203—Market surveys; Market polls
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Economics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Game Theory and Decision Science (AREA)
Abstract
The invention discloses a medicine sales prediction method based on a BP neural network and ARIMA combined model. The problem that a single prediction method based on a traditional research method or an artificial neural network is of low prediction accuracy is solved. First, the historical annual sales of a medicine is predicted using an ARIMA model, and the linear law information is contained in the prediction error of the ARIMA model; then, the error of the ARIMA model is predicted using a BP neural network, and the nonlinear law is contained in the prediction result of the BP neural network; and finally, the prediction result of the ARIMA model and the prediction result of the BP neural network are added up to get the predicted value of the combined prediction model. The method can overcome the defect of a time series method in medicine sales prediction to a large extent and improve the accuracy of medicine sales prediction significantly. The method can be used to predict the medicine sales properly, and can be used as a medicine sales prediction method in the future. The medicine sales prediction process can be implemented simply through Eviews software. The method is of high practicability, and is easy to popularize and apply.
Description
Technical field
The present invention relates to computer technology in pharmaceuticals industry realm of sale, the prediction of the medical sales volume of a kind of ARIMA model based on mixing and artificial nerve network model.
Background technology
Medical market is very significant considering that as the important component part of Chinese society market economy, the sustainable development for national economy.In medicine marketing, medicine dispensing is an important content therein, updates dispensing quantitative forecasting technique, improves precision of prediction significant to pharmaceutical manufacturer.By the forecast analysis to medicine sales, pharmaceutical manufacturer can more reasonably determine medicine distribution kind and quantity, thus reduce entreprise cost, increases the benefit, and is that enterprise marches toward informationization further；Pharmaceutical manufacturer can grasp the sales situation of each quasi drugs, and then the contradiction between balanced linkage pharmacy and enterprise by medicine Method for Sales Forecast；The medicine if chain pharmacy can not reasonably be provided and delivered, it is ensured that the demand of sale, not only can bring medicine short supply to cause the loss of sales opportunnities, make sales achievement decline；And also can injure the relation between consumer and enterprise, affect the prestige of enterprise and the occupation rate of the market share；If when medicine sales amount is the best, and the medicine caused overstocks phenomenon also can be increasingly severe, can increase the expense needed for drug stock, then affecting business capital flowing, increase enterprise's load, this is all the most disadvantageous to enterprise.Generally speaking, medicine sales prediction is the best method solving above-mentioned contradiction exactly, by feasible Forecasting Methodology, draw ideal predicting the outcome, then corresponding dispensing medicine scheme is made according to predicting the outcome, it is that all departments of enterprise can well coordinate, reduces unnecessary expenditures, reach profit maximization and be very important.
At present, there is the Forecasting Methodology of multiple sales volume, such as time series, regression analysis, gray prediction, neutral net, fuzzy technology and genetic algorithm etc., in actual sales forecasting, the most all achieve ideal effect.Wherein, using most is time series method, the most only by sell by the history medicine moon Changing Pattern of sales volume directly predict future sell sales volume by the medicine moon, its advantage is simple, it is easy to grasp, can fully use each item data of former seasonal effect in time series, calculate speed fast, model parameter is had the ability being dynamically determined.Therefore, if directly finding the law of development of the history medicine moon amount of selling, carrying out the amount of the selling prediction of the medicine moon on this basis and will result in the problem that precision of prediction is the highest.
Accordingly, it would be desirable to a kind of Forecasting Methodology for pharmaceuticals industry medicine sales volume.
Summary of the invention
An object of the present invention is the Forecasting Methodology proposing a kind of medical sales volume based on mixed model；The two of the purpose of the present invention are the prognoses systems proposing a kind of medical sales volume based on mixed model.
An object of the present invention is achieved through the following technical solutions:
The Forecasting Methodology of a kind of based on mixed model the medical sales volume that the present invention provides, comprises the following steps:
Set up ARIMA model；
Obtain medicine sales volume historical data sequence and be input to ARIMA model prediction and obtain ARIMA model predictive error and ARIMA and predict the outcome；
Set up BP neural network model；
ARIMA model predictive error is input to BP neural network model be predicted obtaining BP neural network prediction result；
Predict the outcome ARIMA to superpose with BP neural network prediction result and predicted the outcome.
Further, described ARIMA model is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function.
Further, described difference number of times d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be the difference number of times d of medicine sales volume historical data sequence.
Further, described autoregressive coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw the autocorrelogram through d differentiated medicine sales volume historical data sequence；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p.
Further, described rolling average coefficient q is obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw the partial autocorrelation figure through d differentiated medicine sales volume historical data sequence；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
Further, described ARIMA model predictive error is obtained by following steps:
According to difference number of times d, autoregression item number p and rolling average item number q determine medicine sales volume historical data sequence have linear rule part forecast model ARIMA (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
Further, described BP neural network model uses three layers of BP neural network model.
The two of the purpose of the present invention are achieved through the following technical solutions:
The prognoses system of a kind of based on mixed model the medical sales volume that the present invention provides, it is characterised in that: include medicine sales volume historical data sequence acquisition module, ARIMA model module, BP neural network model module, predict the outcome module；
Described medicine sales volume historical data sequence acquisition module, is used for obtaining medicine sales volume historical data sequence；
Described ARIMA model module, is used for setting up ARIMA model；And medicine sales volume historical data sequence inputting to ARIMA model prediction acquisition ARIMA model predictive error and ARIMA are predicted the outcome；
Described BP neural network model module, is used for setting up BP neural network model；And ARIMA model predictive error is input to BP neural network model be predicted obtain BP neural network prediction result；
The described module that predicts the outcome, is predicted the outcome for predicting the outcome to superpose with BP neural network prediction result by ARIMA.
Further, described ARIMA model module is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function；
Described difference number of times d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be the difference number of times d of medicine sales volume historical data sequence；
Described autoregressive coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw the autocorrelogram through d differentiated medicine sales volume historical data sequence；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p；
Described rolling average coefficient q is obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw the partial autocorrelation figure through d differentiated medicine sales volume historical data sequence；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
Further, described ARIMA model predictive error is obtained by following steps:
According to difference number of times d, autoregression item number p and rolling average item number q determine medicine sales volume historical data sequence have linear rule part forecast model ARIMA (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
Owing to have employed technique scheme, present invention have the advantage that:
A kind of based on BP neutral net and ARIMA built-up pattern the medicine Method for Sales Forecast method that the present invention provides；Solve the problem that individual event Forecasting Methodology precision of prediction based on tradition research method or artificial neural network is low.This method is first with the history annual turnover of ARIMA model prediction quasi drugs, it is that its linear rule information has been included in the forecast error of ARIMA model, then by the error of BP neural network prediction ARIMA model so that it is non-linear rule is included in the predicting the outcome of BP neutral net.The predictive value obtaining combination forecasting finally it is added with the prediction with BP neutral net that predicts the outcome of ARIMA；This method can overcome the time series method defect when predicting medicine sales volume largely, significantly improves medicine Method for Sales Forecast precision；Can preferably predict medicine sales volume, can be as following medicine Method for Sales Forecast method；Can be practical, it is simple to popularization and application by its prediction medicine sales volume process of Eviews software simple realization.
Other advantages, target and the feature of the present invention will be illustrated to a certain extent in the following description, and to a certain extent, will be apparent to those skilled in the art based on to investigating hereafter, or can be instructed from the practice of the present invention.The target of the present invention and other advantages can be realized by description below and obtain.
Accompanying drawing explanation
The accompanying drawing of the present invention is described as follows.
Fig. 1 is ARIMA-ANN mixed model schematic diagram.
Fig. 2 be medicine Method for Sales Forecast implement process.
Fig. 3 is the sales volume curve chart (unit: ten thousand) of 1980-1984 streptomycin.
Fig. 4 is first difference curve chart.
Fig. 5 is the autoregression figure on the 1-20 rank of sales data sequence after first difference.
Detailed description of the invention
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
Embodiment 1
As it can be seen, the Forecasting Methodology of a kind of based on mixed model the medical sales volume of the present embodiment offer, comprise the following steps:
Set up ARIMA model；
Obtain medicine sales volume historical data sequence and be input to ARIMA model prediction and obtain ARIMA model predictive error and ARIMA and predict the outcome；
Set up BP neural network model；
ARIMA model predictive error is input to BP neural network model be predicted obtaining BP neural network prediction result；
Predict the outcome ARIMA to superpose with BP neural network prediction result and predicted the outcome.
Described ARIMA model is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function.
Described difference number of times d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be the difference number of times d of medicine sales volume historical data sequence.
Described autoregressive coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw the autocorrelogram through d differentiated medicine sales volume historical data sequence；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p.
Described rolling average coefficient q is obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw the partial autocorrelation figure through d differentiated medicine sales volume historical data sequence；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
Described ARIMA model predictive error is obtained by following steps:
According to difference number of times d, autoregression item number p and rolling average item number q determine medicine sales volume historical data sequence have linear rule part forecast model ARIMA (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
Described BP neural network model uses three layers of BP neural network model.
The acf auto-correlation function that the present embodiment provides is calculating difference function diff in R language, is used for calculating vector or the finite difference of array (by row)；
The medicine concrete processing procedure of sales volume historical data that the present embodiment provides is as follows:
source<-c(110.672,113.685,128.301,85.935,117.725,131.058,126.537,106.203,
143.946,133.739,115.436,97.001,111.753,112.773,86.139,76.955,98.684,103.113,110.837,70.918)
Sale <-ts (source, frequency=4, start=c (1980,1))
Use the time series of ts function productive experiment data, from the first quarter in 1980, annual 4 season
Salediff1 <-diff (sale, differences=1)
plot.ts(salediff1)
Salediff2 <-diff (sale, differences=2)
plot.ts(salediff2)
Salediff3 <-diff (sale, differences=3)
plot.ts(salediff3)
Use diff function that time series data carries out 1,2,3 difference respectively, and draw time-serial position figure after difference, select difference number of times more smoothly
Acf (salediff1, lag.max=20)
Pacf (salediff1, lag.max=20)
By acf function, pacf function tries to achieve autocorrelogram and partial autocorrelation figure after difference respectively, determines suitable delayed number of times, so that it is determined that autocorrelation coefficient and PARCOR coefficients
Salearima <-arima (sale, order=c (0,1,1))
Salearimaforecast <-forecast.Arima (salearima, h=4, level=c (95.5))
After trying to achieve the parameters of arima model, set up arima model by arima function, be then predicted by forecast.Arima function, try to achieve predicting the outcome of Arima model
Net <-newff (n.neurons=c (4,8,2,1), learning.rate.global=1e-2, momentum.global=0.5, error.criterium=" LMS ", Stao=NA, hidden.layer=" tansig ", output.layer=" purelin ", met hod=" ADAPTgdwm ")
Result <-train (net, data1 [, 1:4], data1 [, 5], error.criterium=" LMS ", report=TRUE, sh ow.step=100, n.shows=5)
y<-sim(result$net,data1[5:8,])
Set up three layers of BP neural network model by newff function, then use data that the model set up is trained by train function, try to achieve predicting the outcome of BP neutral net finally by sim function.
The present embodiment additionally provides the prognoses system of a kind of medical sales volume based on mixed model, and including medicine sales volume historical data sequence acquisition module, ARIMA model module, BP neural network model module, predict the outcome module；
Described medicine sales volume historical data sequence acquisition module, is used for obtaining medicine sales volume historical data sequence；
Described ARIMA model module, is used for setting up ARIMA model；And medicine sales volume historical data sequence inputting to ARIMA model prediction acquisition ARIMA model predictive error and ARIMA are predicted the outcome；
Described BP neural network model module, is used for setting up BP neural network model；And ARIMA model predictive error is input to BP neural network model be predicted obtain BP neural network prediction result；
The described module that predicts the outcome, is predicted the outcome for predicting the outcome to superpose with BP neural network prediction result by ARIMA.
Described ARIMA model module is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function；
Described difference number of times d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be the difference number of times d of medicine sales volume historical data sequence；
Described autoregressive coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw the autocorrelogram through d differentiated medicine sales volume historical data sequence；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p；
Described rolling average coefficient q is obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw the partial autocorrelation figure through d differentiated medicine sales volume historical data sequence；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
Described ARIMA model predictive error is obtained by following steps:
According to difference number of times d, autoregression item number p and rolling average item number q determine medicine sales volume historical data sequence have linear rule part forecast model ARIMA (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
Embodiment 2
A kind of based on BP neutral net and ARIMA built-up pattern the medicine Method for Sales Forecast method that the present embodiment provides.The method is for only by the problem that individual event Forecasting Methodology precision of prediction based on tradition research method or artificial neural network is low.R language can be used simply to realize as it is shown in figure 1, the present embodiment realizes above-mentioned prediction process.
The technical scheme realizing project of the present invention makes: first, by the steady former medicine sales volume sequence of diff difference function in R language pack, determine difference number of times d, then use the acf auto-correlation function in R language pack to determine the autoregressive coefficient p of former medicine sales volume sequence, finally use the pacf partial autocorrelation function in R language pack to determine the rolling average item number of former medicine sales volume sequence.So that it is determined that ARIMA (p, d, q) model of former medicine sales volume time series linear rule part.Then for ARIMA, (p, d, q) predicted portions of forecast model uses BP neural network model to be predicted, and present invention uses three layers of BP neural network model, i.e. comprises only the BP neural network model of a hidden layer.Finally by ARIMA (p, d, q) model predict the outcome and predicting the outcome of three layers of BP neural network model be added, obtain finally predicting the outcome of former medicine sales volume.Described method to be embodied as step as follows:
Determine the difference number of times d of former medicine sales data sequence p_data.
1. first by R language pack providing diff difference function former medicine sales data sequence p_data is carried out first difference, differentiated time series chart d_1 is drawn.
The most former medicine sales data sequence can be carried out second order difference, three difference, and make differentiated curve chart.
The most substantially after carrying out 3 difference, the most jiggly time series will become the most steady, relatively carry out the curve chart after multi-difference, select a stable differentiated time series chart of comparison, corresponding difference number of times to be the difference number of times d of former medicine sales sequence p_data.
Determine the autoregressive coefficient p of former medicine sales data sequence.
1. use the acf auto-correlation function that R language pack provides, and to arrange lag order lag be 20, draws the 1-20 rank autocorrelogram acf_p through d time above differentiated medicine sales data sequence.
2. the autocorrelogram acf_p tried to achieve is observed behind p rank no longer beyond confidence border, so that it is determined that autocorrelation coefficient p.
Determine the rolling average coefficient q of former medicine sales data sequence.
1. use the pacf partial autocorrelation function that R language pack provides, and to arrange lag order lag be 20, draws the 1-10 rank partial autocorrelation figure pacf_p through d time above differentiated medicine sales data sequence.
2. the partial autocorrelation figure pacf_p tried to achieve is observed behind q rank no longer beyond confidence border, so that it is determined that rolling average coefficient q.
(p, d, q) predict former medicine sales data sequence p_data, and tries to achieve ARIMA (p, d, forecast error value sequence w_data q) to use ARIMA.
1. the difference number of times d determined according to previous step, autoregression item number p and rolling average item number q, determine former medicine sales data sequence have linear rule part forecast model ARIMA (p, d, q).
2. (p, d, q) model prediction former medicine sales data sequence p_data, tries to achieve predictive value sequence f_data to use the ARIMA that tries to achieve.
3. use former medicine sales data sequence p_data to deduct predictive value sequence f_data tried to achieve, obtain ARIMA (p, d, forecast error sequence w_data q).
(p, d, q) forecast error value sequence w_data obtain predictive value sequence bp_data of BP neutral net to use three layers of BP Neural Network model predictive ARIMA obtained in the previous step.
1. according to the forecast error sequence tried to achieve, determining n input item of three layers of BP neural network model, the method that wherein determines is every 5 network input items as three layers of BP neural network model of forecast error sequence.
2. the BP neural network algorithm using the 5-8-1 network structure of Lai Wenboge-Markwardt (L-M algorithm) is predicted.
F_data and bp_data obtained by previous step is added, and obtains the predictive value pf_data of former medicine sales data sequence p_data.
Embodiment 3
This enforcement sample have employed China Heilongjiang Province 1980-1985 streptomycin sales volume quarterly, wherein uses 1980-1984 streptomycin sales volume quarterly as training data, and within 1985, streptomycin sales volume quarterly is as test data.
Sample data is as shown in table 1, table 1. sample data (unit: ten thousand)
| 1 | 2 | 3 | 4 | |
| 1980 | 110.672 | 113.685 | 128.301 | 85.935 |
| 1981 | 117.725 | 131.058 | 126.537 | 106.203 |
| 1982 | 143.946 | 133.739 | 115.436 | 97.001 |
| 1983 | 111.753 | 112.773 | 86.139 | 76.955 |
| 1984 | 98.684 | 103.113 | 110.837 | 70.918 |
| 1985 | 98.017 | 95.180 | 96.070 | 83.140 |
Concrete prediction steps is as follows:
Determine ARIMA (p, d, q) parameter p of model, d, q, and the ARIMA model prediction streptomycin sales volume quarterly in 1985 that use is tried to achieve.
1. R language is used to be loaded into 1980-1984 streptomycin sales volume quarterly, and m-sales volume curve chart 3 when drawing.Then it is carried out first difference, and draw the Fig. 4 after first difference, now find after first difference that sales volume curve becomes comparison steady, thus may determine that difference number of times d is 1.Therefore sales data sequence salediff1 after first difference is obtained.
2. sales data sequence salediff1 after first difference carries out auto-regressive analysis, and acf Autoregressive Functions acf (salediff1, lag.max=20) using R language to provide tries to achieve the autoregression Fig. 5 on 0-20 rank.As can be seen from Figure 5 all values behind 0 rank is all without departing from confidence border, it is possible to determine that autoregressive coefficient p is 0.
3. sales data sequence salediff1 after first difference carries out inclined auto-regressive analysis, and pacf partial autocorrelation function pacf (salediff1, lag.max=20) using R language to provide tries to achieve the partial autocorrelation figure on 1-10 rank.Then all values behind 1 rank is obtained all without departing from confidence border, it is possible to determine that inclined autoregressive coefficient q is 1.
4. according to the p tried to achieve above, d, q value can set up ARIMA (0,1,1) forecast model, and (p, d, q) forecast model prediction sales data source, specifically can realize in R language then to use ARIMA.
Finally illustrate is, above example is only in order to illustrate technical scheme and unrestricted, although the present invention being described in detail with reference to preferred embodiment, it will be understood by those within the art that, technical scheme can be modified or equivalent, without deviating from objective and the scope of the technical program, it all should be contained in the middle of scope of the presently claimed invention.
Claims (10)
1. the Forecasting Methodology of a medical sales volume based on mixed model, it is characterised in that: comprise the following steps:
Set up ARIMA model；
Obtain medicine sales volume historical data sequence and be input to ARIMA model prediction acquisition ARIMA model predictive error and ARIMA
Predict the outcome；
Set up BP neural network model；
ARIMA model predictive error is input to BP neural network model be predicted obtaining BP neural network prediction result；
Predict the outcome ARIMA to superpose with BP neural network prediction result and predicted the outcome.
2. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described ARIMA
Model is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function.
3. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described difference time
Number d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be
The difference number of times d of medicine sales volume historical data sequence.
4. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described autoregression
Coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw through d differentiated medicine sales volume historical data sequence
Autocorrelogram；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p.
5. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described movement is put down
All coefficient q are obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw through d differentiated medicine sales volume historical data sequence
The partial autocorrelation figure of row；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
6. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described ARIMA
Model predictive error is obtained by following steps:
Determine that medicine sales volume historical data sequence has linearly according to difference number of times d, autoregression item number p and rolling average item number q
The forecast model ARIMA of rule part (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
7. the Forecasting Methodology of medical sales volume based on mixed model as claimed in claim 1, it is characterised in that: described BP god
Three layers of BP neural network model are used through network model.
8. the prognoses system of a medical sales volume based on mixed model, it is characterised in that: include medicine sales volume historical data sequence
Acquisition module, ARIMA model module, BP neural network model module, predict the outcome module；
Described medicine sales volume historical data sequence acquisition module, is used for obtaining medicine sales volume historical data sequence；
Described ARIMA model module, is used for setting up ARIMA model；And by medicine sales volume historical data sequence inputting to ARIMA
Model prediction obtains ARIMA model predictive error and ARIMA predicts the outcome；
Described BP neural network model module, is used for setting up BP neural network model；And ARIMA model predictive error is inputted
It is predicted obtaining BP neural network prediction result to BP neural network model；
The described module that predicts the outcome, is predicted the outcome for predicting the outcome to superpose with BP neural network prediction result by ARIMA.
9. the prognoses system of medical sales volume based on mixed model as claimed in claim 8, it is characterised in that: described ARIMA
Model module is set up by following steps:
The autoregressive coefficient p of medicine sales volume historical data sequence is determined by acf auto-correlation function；
Difference number of times d is determined by diff difference function steady medicine sales volume historical data sequence；
The rolling average item number q of medicine sales volume historical data sequence is determined by pacf partial autocorrelation function；
Described difference number of times d is obtained by following steps:
Use diff difference function that medicine sales volume historical data sequence is carried out first difference, obtain differentiated time series chart；
Medicine sales volume historical data sequence is carried out second order difference and three difference, and obtains differentiated curve chart；
Relatively carry out the curve chart after multi-difference, select stable differentiated time series chart, corresponding difference number of times to be
The difference number of times d of medicine sales volume historical data sequence；
Described autoregressive coefficient p is obtained by following steps:
Use acf auto-correlation function, and lag order is set, draw through d differentiated medicine sales volume historical data sequence
Autocorrelogram；
Behind p rank, no longer exceed confidence border according to autocorrelogram, determine autoregressive coefficient p；
Described rolling average coefficient q is obtained by following steps:
Use pacf partial autocorrelation function, and lag order is set, draw through d differentiated medicine sales volume historical data sequence
The partial autocorrelation figure of row；
Behind q rank, no longer exceed confidence border according to partial autocorrelation figure, determine rolling average coefficient q.
10. the prognoses system of medical sales volume based on mixed model as claimed in claim 8, it is characterised in that: described ARIMA
Model predictive error is obtained by following steps:
Determine that medicine sales volume historical data sequence has linearly according to difference number of times d, autoregression item number p and rolling average item number q
The forecast model ARIMA of rule part (p, d, q)；
(p, d, q) model prediction medicine sales volume historical data sequence, try to achieve predictive value sequence to use ARIMA；
Use medicine sales volume historical data sequence to deduct the predictive value sequence tried to achieve, obtain ARIMA (p, d, forecast error sequence q).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610264506.2A CN105976199A (en) | 2016-04-26 | 2016-04-26 | Medicine sales prediction method and medicine sales prediction system based on hybrid model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610264506.2A CN105976199A (en) | 2016-04-26 | 2016-04-26 | Medicine sales prediction method and medicine sales prediction system based on hybrid model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105976199A true CN105976199A (en) | 2016-09-28 |
Family
ID=56993956
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610264506.2A Pending CN105976199A (en) | 2016-04-26 | 2016-04-26 | Medicine sales prediction method and medicine sales prediction system based on hybrid model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105976199A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274209A (en) * | 2017-05-18 | 2017-10-20 | 北京京东尚科信息技术有限公司 | The method and apparatus for predicting advertising campaign sales data |
| CN107292672A (en) * | 2017-07-05 | 2017-10-24 | 上海数道信息科技有限公司 | System and method for is realized in a kind of catering industry sales forecast |
| CN107767191A (en) * | 2017-12-05 | 2018-03-06 | 广东技术师范学院 | A kind of method based on medical big data prediction medicine sales trend |
| CN108074003A (en) * | 2016-11-09 | 2018-05-25 | 北京京东尚科信息技术有限公司 | Predictive information method for pushing and device |
| CN108196957A (en) * | 2017-12-28 | 2018-06-22 | 福州大学 | A kind of host load prediction method under cloud environment |
| CN108876421A (en) * | 2017-05-09 | 2018-11-23 | 北京京东尚科信息技术有限公司 | A kind of method and system for predicting commodity dynamic sales volume |
| CN108985808A (en) * | 2017-06-01 | 2018-12-11 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus carrying out commodity selection |
| CN109034905A (en) * | 2018-08-03 | 2018-12-18 | 四川长虹电器股份有限公司 | The method for promoting sales volume prediction result robustness |
| CN109961315A (en) * | 2019-01-29 | 2019-07-02 | 河南中烟工业有限责任公司 | A kind of monthly Method for Sales Forecast method of cigarette based on nonlinear combination model |
| CN109978306A (en) * | 2017-12-28 | 2019-07-05 | 北京京东尚科信息技术有限公司 | Data processing method and device |
| CN110045771A (en) * | 2019-04-19 | 2019-07-23 | 淮阴工学院 | A kind of fishpond water quality intelligent monitor system |
| CN110415013A (en) * | 2019-06-12 | 2019-11-05 | 河海大学 | A kind of combination forecasting method of net about vehicle trip requirements in short-term |
| CN110555578A (en) * | 2018-06-01 | 2019-12-10 | 北京京东尚科信息技术有限公司 | sales prediction method and device |
| CN112001740A (en) * | 2020-06-19 | 2020-11-27 | 南京理工大学 | Combined prediction method based on adaptive neural network |
| CN113191559A (en) * | 2021-05-08 | 2021-07-30 | 苏州瑞城电力科技有限公司 | Improved neural network short-term resident load prediction method based on autoregressive selection |
| CN113837814A (en) * | 2021-11-01 | 2021-12-24 | 北京字节跳动网络技术有限公司 | Method and device for predicting quantity of released resources, readable medium and electronic equipment |
| JP6995233B1 (en) * | 2021-04-06 | 2022-01-14 | エンサイス株式会社 | Information processing systems, computer systems and programs |
-
2016
- 2016-04-26 CN CN201610264506.2A patent/CN105976199A/en active Pending
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074003A (en) * | 2016-11-09 | 2018-05-25 | 北京京东尚科信息技术有限公司 | Predictive information method for pushing and device |
| CN108074003B (en) * | 2016-11-09 | 2021-11-12 | 北京京东尚科信息技术有限公司 | Prediction information pushing method and device |
| CN108876421A (en) * | 2017-05-09 | 2018-11-23 | 北京京东尚科信息技术有限公司 | A kind of method and system for predicting commodity dynamic sales volume |
| CN107274209A (en) * | 2017-05-18 | 2017-10-20 | 北京京东尚科信息技术有限公司 | The method and apparatus for predicting advertising campaign sales data |
| CN108985808A (en) * | 2017-06-01 | 2018-12-11 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus carrying out commodity selection |
| CN107292672A (en) * | 2017-07-05 | 2017-10-24 | 上海数道信息科技有限公司 | System and method for is realized in a kind of catering industry sales forecast |
| CN107767191A (en) * | 2017-12-05 | 2018-03-06 | 广东技术师范学院 | A kind of method based on medical big data prediction medicine sales trend |
| CN108196957A (en) * | 2017-12-28 | 2018-06-22 | 福州大学 | A kind of host load prediction method under cloud environment |
| CN109978306A (en) * | 2017-12-28 | 2019-07-05 | 北京京东尚科信息技术有限公司 | Data processing method and device |
| CN110555578A (en) * | 2018-06-01 | 2019-12-10 | 北京京东尚科信息技术有限公司 | sales prediction method and device |
| CN110555578B (en) * | 2018-06-01 | 2024-04-16 | 北京京东尚科信息技术有限公司 | Sales prediction method and device |
| CN109034905A (en) * | 2018-08-03 | 2018-12-18 | 四川长虹电器股份有限公司 | The method for promoting sales volume prediction result robustness |
| CN109961315A (en) * | 2019-01-29 | 2019-07-02 | 河南中烟工业有限责任公司 | A kind of monthly Method for Sales Forecast method of cigarette based on nonlinear combination model |
| CN110045771A (en) * | 2019-04-19 | 2019-07-23 | 淮阴工学院 | A kind of fishpond water quality intelligent monitor system |
| CN110415013A (en) * | 2019-06-12 | 2019-11-05 | 河海大学 | A kind of combination forecasting method of net about vehicle trip requirements in short-term |
| CN112001740A (en) * | 2020-06-19 | 2020-11-27 | 南京理工大学 | Combined prediction method based on adaptive neural network |
| JP6995233B1 (en) * | 2021-04-06 | 2022-01-14 | エンサイス株式会社 | Information processing systems, computer systems and programs |
| CN113191559A (en) * | 2021-05-08 | 2021-07-30 | 苏州瑞城电力科技有限公司 | Improved neural network short-term resident load prediction method based on autoregressive selection |
| CN113191559B (en) * | 2021-05-08 | 2024-04-19 | 苏州瑞城电力科技有限公司 | Improved neural network short-term resident load prediction method based on autoregressive selection |
| CN113837814A (en) * | 2021-11-01 | 2021-12-24 | 北京字节跳动网络技术有限公司 | Method and device for predicting quantity of released resources, readable medium and electronic equipment |
| CN113837814B (en) * | 2021-11-01 | 2024-08-27 | 北京字节跳动网络技术有限公司 | Method and device for predicting quantity of released resources, readable medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105976199A (en) | Medicine sales prediction method and medicine sales prediction system based on hybrid model | |
| Deepu et al. | Supply chain digitalization: An integrated MCDM approach for inter-organizational information systems selection in an electronic supply chain | |
| Lukyanova et al. | Scenario method of strategic planning and forecasting the development of the rural economy in agricultural complex | |
| Azanha et al. | Agile project management with Scrum: A case study of a Brazilian pharmaceutical company IT project | |
| Şen et al. | Evaluation and pre-allocation of operators with multiple skills: A combined fuzzy AHP and max–min approach | |
| Lan et al. | Trends in sustainable logistics in major cities in China | |
| CN110392899A (en) | The dynamic feature selection generated for model | |
| CN104008428B (en) | Service of goods requirement forecasting and resource preferred disposition method | |
| Odeh et al. | The Biobank Economic Modeling Tool (BEMT): Online financial planning to facilitate biobank sustainability | |
| Vorotnikov et al. | Economic aspects of the development of the" digital agriculture" system. | |
| Kees et al. | An optimization model for managing the drug logistics process in a public hospital supply chain integrating physical and economic flows | |
| Cuong et al. | Application of ICT in Logistics and Supply Chain in post-Covid-19 economy in Vietnam | |
| Tyukhtenko et al. | Economic and mathematic models for staff planning at enterprises of all ownership forms | |
| Arantes et al. | Application of a purchasing portfolio model to define medicine purchasing strategies: An empirical study | |
| Azzaro-Pantel | New product development and supply chains in the pharmaceutical industry | |
| Xue et al. | Order acceptance and scheduling in the instant delivery system | |
| Fani et al. | Implementation framework for PLM: a case study in the fashion industry | |
| Aláč | Decision making and its importance in production planning within the woodprocessing company, respectively in the whole supply chain | |
| Fradinata et al. | Forecasting determinant of cement demand in Indonesia with artificial neural network | |
| Paulo et al. | Assessment of biomass supply chain design and planning using discrete-event simulation modeling | |
| Kembero et al. | Determinants of procurement performance in public entities in Kenya: a case of Kenya school of law | |
| Liu | Study on the Influence of Deep Learning and Artificial Intelligence on Transportation and Mobility Industry and Corporations | |
| Yu et al. | An effective multi-part dedicated flow-line reconfiguration model considering the optimal selection of machining process path and machines | |
| Schrepel et al. | The End of Average: Deploying Agent-Based Modeling to Antitrust | |
| Bansal et al. | The efficiency analysis of two-stage dynamic interval DEA model incorporating cumulative prospect theory: an application to Indian airlines |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160928 |