Prediction of ARIMA Model Based on Season Effects on New Employment Number of Towns in Guizhou Province

1. Introduction

Guizhou Province offensive to eliminate poverty, improve people’s livelihood, and gradually realize common prosperity, is the essential requirements of socialism, and is also an important mission of the Chinese Communist Party. This article is based on the three major principles built in the evaluation of happiness indicators, including regional, representative and operability [1], for the actual situation of Guizhou Province, the achievements of the implementation of the strategy of poverty, from urban employment growth Angle analysis has been developing in Guizhou in recent years. Liu Pan combined with Guizhou Forestry Promoting the Advantages of Poverty Alleviation and Taking Measures, I proposed the Countermeasures for the Work of Forestry Poverty Alleviation [3]. The difference between each stage is considered to be in terms of ice, causing the problems in the presence of poverty, and is constantly changing, and it is necessary to accurately identify the steering and contradictions [4].

Guizhou province has steadily launched in the employment of urban residents, and the vast majority of urban residents are guaranteed. The study found that urban development enabling the surrounding economy underdeveloped, Guizhou Province, poverty-poverty attention should be moved from the countryside [5]. Li Ru is based on gray correlation analysis, pointing out that the changes in urban new employment population have certain periodic [6]. Wujiang believes that the number of new employment in urban and towns is affected by many local economic development, industrial structure, and education levels [7]. The number of new employment can be more intuitive reflect the progress of the provincial provinces of Guizhou Province, so the prediction of the new employment will provide a reference for future povertygrams. Forecast is the basis and basis of decision making, predicting new employment people, can provide certain help for the implementation of relevant policy measures, so the prediction of new employment in Guizhou Province is meaningful and necessary.

2. Theoretical Basic

AriMA Model

ARIMA Model (Autoregressive Integrated Moving Average Model), differential integrated mobile average self-return model, also known as integrated mobile The average self-regression model is one of the time series prediction analysis methods [8]. ARIMA (P, D, Q), Ar is “self-return”, P is the number of self-registration items; MA is “sliding average”, Q is the number of sliding average, D is the number of differential times made to make a smooth sequence . The quarter product ARIMA model is arimaComprehensive model and season model. If time sequence y t In addition to the trend variation, there is a more pronounced seasonal change, in order to perform sequences, eliminate the seasonal, and then perform seasonal differential elimination sequences. The steps should be consistent with the season cycle and then build an ARIMA model containing the parameters of the season differential, that is, the seasonal product ARIMA model. ARIMA (p, d, q) × (p, d, q) S

Model is:

φ ) φ L S ) ( 1 D [ ( – L S ) Y T = θ Q [1 θ q ( ) ε T φ P ([ = 1 – φ 1 φ 2 L – –

φ P is a non-season self-return polynomial, and P is the self-return order; ( L ) = 1 – L – θ 2 – ⋯ – θ L Q

is a non-season moving average polynomial, Q is a moving average; θ p ( L = 1 L – L 2 – Q L q is a seasonal self-return polynomial, wherein P is a seasonal self-return step; θ Q ( =

θ 2 L – – θ q is the season moving average polynomial, where Q is the season moving average; ( – ) D is a differential calculator, D is a differential order; ( 1 –

D For the seasonal difference operator, D is the seasonal difference order, and S is the season cycle. 2.2. Determining factor decomposition method The basic idea of ​​determining factor decomposition is: although different sequence fluctuations are triper, butVarious changes in the sequence can be summarized into a comprehensive impact of four major factors [8]: 1) Long-term trend: The effect of this factor can result in a significant long-term trend (increment, Decrease, etc.); 2) Cycle fluctuation: This factor can cause sequences to present repeated cycle fluctuations from low to high, if the observation period is not long enough to transaction day ( DAY); 3) Seasonal change (Season): This factor can cause sequences to present stable cycle fluctuations related to seasonal changes; 4) Random fluctuation (IMMEDIAT): In addition to long-term trends, cycle fluctuations (trading days), seasonal changes, sequences will also be affected by various other factors, and these effects result in a certain random fluctuation. 3. Analysis and Forecast 3.1. Analysis of new employment population Historical data from January 2002 to December 2019 urban new employment The sequence is recorded as Y, from the timing chart ( Fig. 1 ) can be generally seen that its historical trend and fluctuation situation, the horizontal coordinates in the figure, and the ordinate indicates the number of new employment. Figure 1 . TIME Series of Newly-Increased Urban Employment Population from 2002 to 2019 . 2002 to 2019 urban new employment population timing chart The number of new employment in urban new employment has a significant growth trend, and It is affected by seasonal effects in a certain extent. Here is a broad “season”, that is, any event that presents a fixed periodic change can be called “season” effect. From FIG. 2 It can be seen that there is a significant seasonal effect in the number of new employment populations, new March and JuneThe number of people is significantly higher than that of the other month. Overall, the new employment population presents a gradually increased trend and obvious seasonal characteristics, trend of seasonal fluctuations, and obvious seasonal characteristics, seasonal fluctuations, and gradually slowing. Since Guizhou Province, January 2016, the number of new employment people in the urban new employment in the urban new employment, such as this sequence has a significant growth trend, there is a fixed change cycle, so consider adoption Determinative analysis method, which uses commonly used deterministic factors. Figure 2 . Seasonal FluctionS in new Employment Fig. 2 . Seasonal fluctuations in the new employment

Due to the difference in the amount of the indicator, when the levels between each indicator differ, if the original indicator value is directly utilized Analysis, the role of highlighting the high value levels in integrated analysis, reducing the effect of lower numerical levels. In order to ensure the reliability of the results, after the data is completed, the data is further standardized by SPSS software, and the processing steps: SMS data window Site: Analysis – Description Statistics – Description, select the variables to select the alternate box, select Save a standardized score as a variable, click OK, and get the standardized score of each variable in the data window [9]. 3.2. Establish an ARIMA model Due to the prediction angle, the recent value is more than the long-term value to the future, in order to make the model prediction while using more historical data The accuracy is higher, and this paper selection from January 2002 to the monthly data of the number of new employees in 18 years in 2019. All data from 2002 to 2018 are used as samples for establishing an ARIMA model. In 2019, data is used for sample prediction to verify the fitting effect of the model. This article uses R 3.6.3 software and EVIEWS10.0 software to establish models and predict. The establishment process of the ARIMA model is Fig. 3 is shown. Figure 3 . Arima Modeling Flowchart Figure 3 . ARIMA Modeling Flowchart The ARIMA model is established, and there is a pretreatment, model identification, model set, parameter estimate, and model validity verification of time series. 1) Premieving of time series data must simultaneously meet both of the stability and non-white noise to analyze the prediction of the ARIMA model. If the data is non-steady, it can be used to smooth data, if the sequence is white noise, it means that there is no obvious correlation between the values, that is, the past has no effect on future trend development. Therefore, there is no value, so the pretreatment of the data includes two parts: the stationary test and the white noise test. First, it can be observed from the sequence of sequences, from Fig. 1 It can be seen that the number of new employment has a growth trend, and has obvious seasonal fluctuations, data is not stable. It is seen that its stability is inspected with the unit root (see table 1 ), it can be seen that the ADF inspection statistic value is -2.5690, which is greater than 1%, 5%, 10% of the verification level, so There is a unit root that is non-stationary. Table 1 . UNIT Root Test for SEQUENCE Y Table 1 . The unit root test in order to make the sequence change, and the original sequence Y is a first-order differential process to obtain a sequence Dy. From

Fig. 4

It can be seen that the differential sequence DY always fluctuates up and down in a constant value, and the trend is eliminated.It further conducts the unit root inspection (see

Table 2

), it is understood that the original holiday in which the unit root is rejected, and the sequence DY is smooth.

Figure 4

. Timing Diagram of Sequence Dy After Differge

. Sequence diagram

Table 2 . Unit root test for sequence y

. Sequence y’s unit root inspection

Fast-stability test After passing, the sequence DY is also performed, so it makes the self-relevant and partial related map for further investigation, from

Fig. 5

can be seen, at the 6th order , The P value of 12th order, 18th order, 24th order, Q statistics is less than 0.05, so the sequence is a non-white noise sequence. Through the above test, it is understood that DY is a smooth non-white noise sequence, and an ARIMA model can be established.

Figure 5

. AutocorRelation Graph and Partial AutoCorrelation Graph of Sequence Dy

Sequence diagram adjusted from season (

Fig. 6

) can be seen from the fluctuation of the seasonal factor, but autofinus and partial correlation graph (

Fig. 7 ) The cutting and tail tendency are still not obvious, trying to fit the ARIMA model but the effect is poor. When the latency is still larger, AIC and PAIC are still larger, still obviously prominent in the figure, can be considered that there is still a seasonal effect in the sequence of seasonal adjusted, so the short-term correlation between the sequence and the seasonal effect have complex correlation, You cannot simply extract, you can try to fit the product season ArIMA (P, D, Q) × (P, D, Q) S model.

Figure 6

. Timing Diagram of Seasonally Adjusted Sequence DY12

. Sequence diagram of sequence DY12 after seasonal adjustment

FiguR 7

. AUTOCORRELATION AND Partial Autocorrelation Plots of Sequence DY12

Fig. 7

. Self-correlation and partial search of sequence DY12 Since the sequence has been first or two step differentials, D, D is 1, and s is 12. By 7

may be initially determined to be 2, Q is 1. HereIn order to find the optimal fitting model, 0, 1, 2 is selected for P and Q, respectively, and similar, for P and Q, 0, 1, 2 is selected separately. For the screening of the model, first see the parameter significance, if the parameters are not significant, the indexing of the independent parameters should be removed, and the low-order to high-order is sequentially modeled, and finally ARIMA (1, 1, 1) × (0, 1, 1)

. Comparison of two model correlation statistics

The data in Table 3 can be seen, compared to MOD2 AIC guidelines, SCI standards, and HQC guidelines are smaller, and simultaneously equipped R 2 , it is considered to be better, × (2, 1, 1) × (0, 1, ” 1) 12 is the optimal fitting model.

3) Model Validity Verification

To perform an effective test, that is, check whether the test residual is satisfied, the residual sequence test results are

Fig. 8

. The P value of the Q statistic is much greater than 0.05, so the residual sequence is white noise, so the relevant information has been fully extracted by the model, and the model is effective. . 4) Model fitting effect test established ARIMA (2, 1, 1) × (0, 1, 1) 12

model, in 2019 Data is pre-sampledTest, inspected its fitting effect.

Fig. 9 The real value of the new employment population in 2019 was added, and it can be seen that the actual value and the predicted value were 95%. The prediction interval and both bonded. The predicted value of the model and the predictive section are Table 4


It can be seen that although the actual value is in-depth, the relative error is small, the average error is 3.5%, and the prediction accuracy is better. Sustaining, the ARIMA model predicts the monthly data of 2002 to 2019, and the overall fitting effect is seen. Due to the data, the predictive sample range is reduced to April 2003 to December 2019.

Figure 8

. AutocorRelation and Partial Autocorrelation Plots of Residual Series

Figure 8

. Self-correlation and partial correlation diagram of residual sequence

Figure 9

. The deficated value in 2019

Figure 9

. 2019 new employment population data true The deviation between the value and the predicted value

Table 4 . System Resulting Data of Standard Experiment Table 4

. Model prediction results and true value comparison

The ARIMA (2, 1, 1) × (0, 1, 1)

12 model is established by the above steps, and the resulting model diameter is as follows. (

) 1 L



= +

( 1 +




3.3. Prediction and discussion

The ARIMA model applies to short-term predictions, so this article predicts the foreign direct investment in my country in 2020.

Table 5 Figure 10 is the result of the number of new employment in the urban urban town in 2020.

Table 5 . Comparison of Model Prediction Results with real values ​​ Table 5

. Model prediction results and real value comparison

Figure 10

. 2020 urban new employment population forecast results

If it is not affected by emergencies, the point forecast value is the theoretical value of the new employment. Since new coronal pneumonia has been built in early 2020, it is possible to analyze the impact of this important event on employment, which enables predictive results closer to reality. Compared with the actual number of employment in Guizhou, from January to April, Guizhou Province, compared with the predictive value of the ARIMA model (see

Table 6


FIG. 11

To observe and analyze the extent of new crown epidemic and the trend of new employment population under the epidemic.

Table 6

. Comparison of Model Prediction Results with real values ​​ Table 6 . Model forecastResults and true value comparison

Figure 11

. Comparison of the predicted and actual valuesin 2020

Figure 11 . In 2020, the prediction value of the new employment is compared with the actual value

January, new crown pneumonitis has not spread widely The new population of urban employment still maintains a trend of steady growth, and the actual phase difference is small and within the acceporary. In February, the new crown epidemic in the outbreak period, the new employment population has begun to be greatly affected; in Fig. 11 can also be seen in February, the actual value exceeds 95% level. The lower limit of the prediction section. In March, as China’s epidemic is basically controlled, the relative error is reduced from the previous month, and the actual value is close to the lower limit of the prediction section, and there is a certain stable sign in the range of employment conditions. In April, the number of employment and predicted values ​​were only 0.3%, indicating that the impact of new crown epidemic gradually subsided, and my country’s resident employment began to return to the right track. Therefore, the impact of this major random event in the new crown epidemic is staged, temporary, does not change the overall stable trend. This also shows that the forecast results of the forebet ARIMA model are still effective.

4. Conclusion and Prospect


Huang Jing. Research on the method of constructing a happiness index index system [D]: [Master’s thesis]. Dalian: Northeast University of Finance and Economics, 2007: 14-25. [2] Wang Chunguang.Study on Poverty Adventure and Sustainable Development in Guizhou Province [J] .Journal of Guizhou University for Nationalities (Philosophy and Social Sciences), 2018 (3): 39-56. [3] Liu Pan, Zhang Ying. Analysis of the effectiveness of the poverty alleviation of Guizhou Forestry Based on Statistics [J]. Forestry Economy, 2017 (5): 94-97. In ice. The problem of deprotecing the work of poverty in Guizhou Province and countermeasures [J]. Theory and Contemporary, 2018, 411 (8): 18-20. 茂 文, Nie Xuesong. Guizhou poverty alleviation attack, poverty characteristics and countermeasures [J]. Guizhou National Research, 2000 (3): 44-47. [6] Li Yixi. Analysis of Gray Correlation in the Change Trend of Urban Employment Population in my country [J]. Modern Business, 2016 (33): 193-194. [7] Wujiang, Wang Xin. Problems in public employment services and their solutions [J]. City Issues, 2012 (7): 59-64. [8] Easy Danhui, Wang Yan. Application Time Series Analysis [M]. 5th Edition. Beijing: Renmin University of China Press, 2019: 154-188. Zhang Wenhao, Qi Chunwei. SPSS Statistical Analysis Basic Tutorial [M]. Beijing: Higher Education Press, 2011: 89-91

Leave a Reply

Your email address will not be published. Required fields are marked *