Frontiers in Business, Economics and Management 
ISSN: 2766-824X | Vol. 4, No. 1, 2022 

 

6 

Short	Term	Investment	Rating	Forecast	
-- Based on Outlook Attribute Factors 

Qin Guo 

School of Economics and Management, Chongqing University of Posts and Telecommunications, Chongqing, 400065, China 

 

Abstract: The stock market has been in unpredictable, in view of the stock market dynamic forecast and investment prospect 
value quantitative problem, this paper plans to forward segment regression method, with one year for experimental period, 
quarter as the time window, dynamically select the window period investment prospect attribute factor, forecast the investment 
prospect of the next quarter, and through the entropy right Topsis method sorting quantitative investment prospect value, rating 
the stock.The rating is divided into 5 grades and 13 levels, and then the investment portfolio is constructed according to the rating 
level and the asset quality indicators are selected to test. The results show that the dynamic investment rating method based on 
the prospect attribute factor has a certain investment rating prediction role in the short term, and plays a certain reference role 
for investors' decision-making. 

Keywords: Investment prospects, Forward segment regression, Entropy right Topsis, Investment rating forecast. 
 

1. Introduction 
Since the 21st century, China's capital market has been in 

the process of continuous development and improvement. 
According to relevant statistics, by the end of June 2020, the 
circulating market value of Shanghai and Shenzhen stock 
markets is 52.1 trillion yuan, and the number of investor 
accounts is 480 million [1], It can be seen that China's 
investment market and investment group scale are very large. 
Investors are mainly divided into two categories, one is 
institutional investors, the other is retail investors, the old 
number of institutional investors is low, and most of the retail 
investors, retail relative to institutional investors, financial 
knowledge and investment experience, also lack of scientific 
and rational decisions, most retail investors do not have the 
ability of stock data analysis, in the overall information 
disadvantage[2]. At the same time of a large number of 
investors, analysis institutions also emerged. Their stock 
earnings forecasts and investment rating reports for individual 
stocks are important ways for investors, especially retail 
investors, to obtain information. Such data provide investors 
with investment decisions such as buying and selling stocks. 
Investors can get more useful information by reading the 
earnings forecasts and investment ratings in the analysts 
'reports, and by observing the analysts' tracking behavior[3]. 

Stock investment rating is also known as analyst rating, 
which generally refers to the prediction of institutions or 
independent analysts on the future investment surplus of the 
analysis subject, through quantitative factor model, calculate 
scores and issue research reports to provide decision-making 
suggestions for investors. At present, Wind comprehensive 
rating is one of the more mature ratings in China. The release 
of its rating data is in the form of signing contracts with 
investment analysis institutions, which is expensive, and there 
is no clear information about the whole process of rating [4] 
And most companies are not rated. Therefore, the purpose of 
this paper is to quantify the investment prospect value and 
forecast the investment rating through the comprehensive 
public market information and objective market data, so as to 
provide scientific and effective investment decision analysis 

for investors, especially retail investors. Second, because the 
stock market has always been volatile, it is difficult to predict 
it accurately. This paper provides a dynamically adjusted 
quantitative investment foreground value method, which 
dynamically finds the factors related to the investment rating 
by the forward segment regression method, and calculates the 
investment foreground value as the foreground attribute factor. 
Taking the quarter as the time window, the investment 
prospect attribute factor of the previous quarter is used to 
dynamically predict the investment prospect value of the next 
quarter, rather than using fixed factor. factor selection has 
certain universality and elasticity, which is more in line with 
the multiple changing characteristics of the stock market. 

2. Foreground Attribute Factor 
Screening 

This paper uses the forward segmentation regression 
method to fit the factors of the independent variables of the 
next quarter and the W ind of the next quarter and the factors 
of the next quarter to fit the factors in the experimental period, 
and then obtains the prospect attribute factor of the next year. 

2.1. Sample Selection 
In this chapter, it takes manufacturing companies as a 

sample, selects the quarterly financial data of A-share listed 
companies in 2017-2019, excluding ST's ST and * ST 
enterprises, as well as enterprises with missing financial data 
and Wind comprehensive rating data, and obtains a total of 
352 enterprises, data sources and Wind database. 

2.1.1. Factor Set Construction 
Reviewing the historical literature, there are two main 

measures of investment rating. One is to verify the investment 
rating through stock yield, and the other is through the factors 
of business operation. In 1981, Rolf and Banz found through 
empirical research that the portfolio yield with the largest 
stock market value was lower than the portfolio yield with the 
smallest stock market value, confirming for the first time the 
scale effect of the U. S. stock market[5], Domestic scholars 
Chen Can and Zhao Mengxue also found that there is scale 
effect in Chinese stock market through demonstration[6-7]. 



 

7 

For value factors and financial factors and stock returns, both 
at home and abroad scholars have different conclusions, Fama 
and French in 1993 put forward the famous three-factor asset 
pricing model, explain the phenomenon of risk return excess 
return, they are looking for the factor affecting stock yields 
found that the reciprocal of price-to-book ratio, p-earnings 
ratio can explain the difference in most stock returns [8]. The 
business operation status of enterprises will also have an 
impact on stock returns. Edirisinghe found through research 
that the investment value of enterprises is also reflected in 

four aspects: development ability, profitability, operation 
ability and solvency [9]. Pan Heping established a 
comprehensive financial evaluation system of the company's 
internal value from five aspects: solvency, operation ability, 
development ability, profitability and investment return 
ability, so as to study the investment value of the enterprise 
[10]. In this chapter, a total of 25 factors are selected to 
construct independent variable factor candidates. The specific 
factor indicators are shown in the table below. The dependent 
variable is the comprehensive rating in the Wind database. 

 
Table 1. The candidate set of independent variable factors 

Independent variable 
factor type 

Independent variable factor name 

Scale factor 
Total 

market value 
Circulation 

market value 
   

Value factor pe ratio 
Price-to-book 

ratio 
City rate City sales rate  

Growth factor 

Year-on-
year growth 

rate of 
operating 
revenue 

Earnings per 
share growth 

rate 

Year-on-year 
growth rate of 

operating profit 

Year-on-year 
growth rate of cash 

flow generated 
from operating 

activities 

Net profit 
year-on-year 
growth rate 

momentum factor 

Price 
momentum for 

the last 10 
days 

Price 
momentum for 
the last 20 days 

Price 
momentum for 
the last 60 days 

Price 
momentum for the 

last 120 days 
 

Market predictors 
Consistent 

forecast of net 
profit 

Consistently 
forecast earnings 

per share 
   

Profit factor 
Return on 
equity 

rate of return 
on total assets 

earnings per 
share 

net profit rate  

Operation class factor 
inventory 

turnover ratio 
turnover of 

total capital 
   

Debt repayment factor quick ratio 
asset-liability 

ratio 
   

2.2. Foreground Attribute Factor Filter 
2.2.1. Data Preprocessing 

In order to eliminate some extreme value interference, the 
data needs to be depolarized, and the MAD method is selected 
for extreme value processing in this chapter. At the same time, 
because the selected factors have different magnitude, the 
factor data need to be standardized in order to compare the 
data of different magnitude. 

2.2.2. Factor Screening 
1. Introduction to the forward piecewise regression 

methods 
The forward stepwise regression algorithm is a method to 

calculate the best regression coefficient after multiple 
iterations. In each iteration, a certain weight is added or 
reduced, observe the error rate changes, and choose the 
direction with little influence on the error. This method is 
suitable for processing high-dimensional data and can avoid 
multiple collinearity between factors [11] However, in this 
method, we need to go through multiple passes and update the 
factor coefficients to find the factors related to the response 
variable. The specific steps are as follows 

(1) Build the independent variable matrix and the response 
variable matrix𝑋𝑌 

𝑋 𝑥 𝑥 , ,…𝑥
𝑥 ⋯ 𝑥

⋮ ⋱ ⋮
𝑥 ⋯ 𝑥

，𝑌

⎣
⎢
⎢
⎢
⎢
⎡
𝑦
𝑦
.
.
.

𝑦 ⎦
⎥
⎥
⎥
⎥
⎤

          (1) 

 
(2) Data standardization 
The independent variable matrix was normalized, and the 

dependent variable matrix was centralized 
(3) Key parameters of the initialization model 
𝑎 , 𝑎 , … … 𝑎 0𝑎  Initializes the coefficients of all 

feature variables in the model to zero, i. e., where (i =1,2,..., 
m) is the coefficient of the i-th variable.The model residual, e 
is initialized as follows: 

 
_

e y y                  (2) 
 

1

n

i
i

y
y

n

                 (3) 

 
(4) Look for the characteristic variables most related to the 

current residuals. 
𝑥 ∈ 𝑅 First, solve each input feature variable (j=1,2,..., m) 



 

8 

Correlation coefficient with the current residual e, with the 
following formula: 

 
cov( , )

( , )
var[ ] var[ ]

j
j

j

x e
corr x e

x e
 (j =1,2,…,m )     (4) 

 

By trathrough the correlation coefficient with the residue 
of each feature variable, find the variable most associated 
with the current residue, when the correlation coefficient 
between the factor and the residue e is, and has = 
max. 𝑥 𝑥 𝑐𝑜𝑟𝑟 𝑥 , 𝑒 𝑐𝑜𝑟𝑟 𝑥 , 𝑒

1 2{ ( , ), ( , ),..., ( , )}mcorr x e corr x e corr x e  
(5) Update the coefficients and residues of the feature 

variables. 
𝑥 𝑎 𝑎 𝛿 𝛿 𝜀 ∙ sign 𝑐𝑜𝑟𝑟 𝑥 , 𝑒 𝜀 The updated 

coefficients, where =, are arbitrarily small step sizes, for 

example =0.01, update residue =.𝜀𝑒𝑒 𝛿 ⋅ 𝑥  
(6) Repeat steps 4 and 5 until any feature variables are 

associated or minimal to the current residue. 
2. Foreground Attribute Factor Screening 
This chapter depolarity and standardizes the resulting 25 

independent variable factor matrix XtThe W ind composite 
rating with the next quarter is the dependent variable matrix 
Yt+1𝜀Enter the forward segment regression model for factor 
screening, select =0.001 as the iteration step in the algorithm, 
predict the 2019 forward value as an example, fit the 
independent variable factor of the fourth quarter of 2018 and 
the first quarter of 2019 Win d for four times, and select the 
factor with repetition rate of more than 75% is selected as the 
foreground attribute factor in this time window. The resulting 
prospect attribute factor and forward segment regressor 
coefficient are shown in Table 2 below.2 

 

Table 2. Foreground attribute factors and factor coefficients 

 pe ratio 
Price-to-

book ratio 

increase 
rate of 

business 
revenue 

Growth rate of 
net cash flow in 

operating 
activities 

Price momentum 
for the last 120 

days 

Consistently 
forecast earnings 

per share 

Return 
on 

equity 

1812 0.192 0.229 -0.037  -0.003 -0.002 -0.309 
1903 0.751 1.446 -0.109 -5.005 -0.005 -0.053 -0.085 
1906 0.571  -0.47 -0.601  -0.042 -0.249 
1909 1.2105 0.3945 -0.083 -2.621 -0.013 -0.03 -0.5705 
Select 

the rate 
100% 75% 100% 75% 75% 100% 100% 

 
As can be seen from the above table, from the fourth 

quarter of 2018 to the third quarter of 20193, the consistent 
forecast of earnings per share, return on equity, operating 
income growth rate, and p / E ratio probability were all 100%, 
indicating that these three factors showed a high correlation 
with Wind rating during this period. The price / E ratio and 
price / book ratio, the higher the price / E ratio and price / 
book ratio, the higher the score, the lower the rating; the 
operating revenue growth rate, cash flow growth rate of 
operating activities, price momentum, consistent forecast 
earnings per share and return on equity preference, and the 
larger these factors, the lower the rating score, the higher the 
rating grade, the factors selected in the above table are defined 
as the prospect attribute factors in this period. 

3. Investment Outlook Value 
Calculation and Investment Rating 
Forecast 

Since there is no definite function relationship between the 
prospect attribute factor and the investment rating, it is 
necessary to quantify the investment prospect according to 
reflect the investment rating of the prospect attribute factor to 
provide investors with more intuitive decision-making 
suggestions. Entropy right Topsis first calculates the index 
weight according to the entropy weight method, and then 
calculates the score through Topsis. This method makes full 
use of the original data information, and the results can 
accurately reflect the gap between the evaluation schemes, 
and there is no strict limit on the data distribution and sample 
content. 

3.1. Calculation of the Foreground Attribute 
Factor Weight 

Entropy weight method is an objective empowerment 
method, which can avoid the interference of subjectivity to 
the experimental results. Its main idea is to make objective 
weight empowerment according to the variability of each 
index. Generally speaking, the smaller the information 
entropy of a certain index, the greater the degree of variation 
of the index, the greater the influence it can play in the 
comprehensive evaluation, and the greater the corresponding 
weight.The entropy weight method is calculated as follows 

(1) The entropy value of each indicator is calculated, as 
shown in the formula (5) 

 
𝑒 = 𝑘 ∑ 𝑝 𝑙𝑛𝑝 ,   𝑗 1,2, … … 𝑚)      (5) 

 
𝑝

∑
, 𝑘 , 𝑥  represents the j index value of 

each i enterprise, m represents the number of index value, and 
n represents the number of enterprises. 

(2) The entropy weight of each index factor is calculated, 
and the formula is shown in (6)𝑤  

 

𝑤
∑

                    (6) 

 
Through the above steps, the weight set of each index can 

be obtained, which can reflect the objective weight size of 
each index.𝑊The entropy right method has been applied to 
many fields. Chen Xiaoxin uses the entropy right T op sis 
method to select indicators for comprehensive evaluation of 



 

9 

white liquor listed companies. The evaluation results show 
that the objective weight is more effective for the performance 
level of rating enterprises [12]. Zhang Mei et al. applied the 
entropy right method to the science and technology rating 
system reflected in the scientific outlook on development, and 
established a comprehensive evaluation model of science and 
technology [13]. Based on the relevant literature, the entropy 
right Top sis multi-attribute decision analysis method is 

selected to calculate the comprehensive score of each 
enterprise and take this as the prospect value of the enterprise 
to quantify the investment prospect. 

Taking the predicted 2019 foreground value as an example, 
the weights of each foreground attribute factor calculated by 
the selected foreground attribute factors according to the 
above steps are shown in Table 3. 

 
Table 3. Four quarter quarter quarter property for four factors 

 pe ratio 
Price-to-

book ratio 

increase 
rate of 

business 
revenue 

Growth rate of 
net cash flow 
in operating 

activities 

Price 
momentum 
for the last 
120 days 

Consistently 
forecast earnings 

per share 
Return on 

equity 

1812 0.089531 0.051943 0.155783 0.079334 0.463861 0.100514 0.059033 
1903 0.134334 0.054837 0.200067 0.031274 0.267414 0.071451 0.240623 
1906 0.022731 0.047549 0.1931 0.075389 0.370872 0.055013 0.235346 
1909 0.060543 0.041456 0.24107 0.049277 0.285182 0.065667 0.256804 

 
Next, a descriptive statistical analysis of the prospect attribute factor weights from Q18 to Q19 is shown in Table 4 

 
Table 4. Descriptive statistical analysis of foreground attribute factor weights 

 pe ratio 
Price-to-

book ratio 

increase rate of 
business 
revenue 

Growth rate of net cash 
flow in operating 

activities 

Price momentum 
for the last 120 

days 

Consistently 
forecast earnings per 

share 

Return on 
equity 

mean 0.076785 0.048946 0.197505 0.058819 0.346832 0.073161 0.197951 
variance 0.047117 0.005823 0.03495 0.022695 0.090151 0.019465 0.093061 

least value 0.022731 0.041456 0.155783 0.031274 0.267414 0.055013 0.059033 
median 0.075037 0.049746 0.196584 0.062333 0.328027 0.068559 0.237984 

crest value 0.134334 0.054837 0.24107 0.079334 0.463861 0.100514 0.256804 
Observations 4 4 4 4 4 4 4 

 
As can be seen from the above table, the weights of the 

seven foreground attribute factors are relatively stable. 
Combining the weight information of the four quarters, the 

final weight of the most median foreground attribute factor is 
finally selected to get the following table 

 
Table 5. Foreground Attribute Factor Weight 

 pe ratio 
Price-to-

book ratio 

increase 
rate of 

business 
revenue 

Growth rate of 
net cash flow 
in operating 

activities 

Price 
momentum 
for the last 
120 days 

Consistently 
forecast earnings 

per share 
Return on 

equity 

weight 0.076785 0.048946 0.197505 0.058819 0.346832 0.073161 0.197951 

3.2. Investment Outlook Value Is Calculated 
By Rating and Forecast 

The Topsis method was developed in 1981 by Wang 
C.L.And the Yoon K.S.For the first time, it is a sort method 
to approximate the ideal solution[14]The method needs to 
determine the positive and negative ideal solution, positive 
ideal solution for the optimal solution of each factor, negative 
ideal solution for the worst solution of each factor, and then 
calculate the distance between the positive ideal point and the 
negative ideal point, finally calculate the evaluation object to 
the relative proximity of the positive ideal point of the 
comprehensive score, the specific steps are shown in the 
figure below. 

The initial decision matrix is constructed for the set of 
evaluation objects and rating indicators as follows 𝐴
𝑎 , 𝑖 1,2, … … 𝑛𝐶 𝑐 , 𝑗 1,2, … … 𝑚𝑋 

𝑋
𝑥 ⋯ 𝑥

⋮ ⋱ ⋮
𝑥 ⋯ 𝑥

 

𝑥  represents the j index value of each i enterprise, m 
represents the number of index value, and n represents the 
number of enterprises. 

Decision matrix normalization. 
Since there are indicators of different dimensions in the 

decision matrix, in order to eliminate the influence of 
different dimensions on the calculation results, the matrix 
should first be normalized, and the normalized matrix 𝐵
𝐵 𝑛 𝑚 is obtained after processing by formula (7) 

 
 

𝑏     𝑖 1,2 … … 𝑛, 𝑗 1,2 … … 𝑚         (7) 

 
Determine the positive and negative ideal solutions.𝑄 𝑄  

𝑄 𝑏 , 𝑏 , … …

max 𝑏 𝑗 ∈ 𝐽 , min 𝑏 𝑗 ∈ 𝐽 | 𝑖

1,2, … … 𝑛  



 

10 

𝑄 𝑏 , 𝑏 , … …

max 𝑏 𝑗 ∈ 𝐽 , min 𝑏 𝑗 ∈ 𝐽 | 𝑖

1,2, … … 𝑛  

𝐽 𝐽 Among them is the income index, is the cost index. 
The bigger the income index, the better, and the smaller the 
cost index, the better. 

The weight of each index is calculated according to the 
entropy weight method𝑤  

The Eucandean weighted distance between the positive and 
negative ideal solutions is calculated for each evaluation 
object 

 

𝐷 ∑ 𝑤 𝑏 𝑏    𝑖 1,2, … … 𝑛         (8) 

 

𝐷 ∑ 𝑤 𝑏 𝑏    𝑖 1,2, … … 𝑛    (9) 

 
Calculate the relative closeness of the rating object to the 

ideal solution𝐶  
 

𝐶                           (10) 
 
𝐶 Order each rating object according to the size, the larger 

the value, the higher the evaluation object score.𝐶  
According to the weights of each factor in Table 3.3, 

according to the above steps, each factor is divided into cost 
indicators and benefit indicators. According to Table 2, the 
price / E ratio and book ratio preference direction are cost 
indicators; operating income growth rate, cash flow growth 
rate in operating activities, price momentum in the past 120 
days, consistent forecast earnings per share and return on 
equity preference direction are benefit indicators. Then, the 
set of positive and negative ideal solutions is constructed and 
the distance from 352 enterprises to positive ideal and 
negative ideal solutions is calculated respectively. Finally, the 
proximity of each enterprise to the positive ideal solution 
from the fourth quarter of 2018 to the third quarter of 2019, 
that is, the comprehensive evaluation score of the fourth 
quarter of 18 is taken as the prospect value of the first quarter 
of 19, and so on, the cumulative distribution of the 
comprehensive evaluation score of four quarters, namely the 
investment prospect value, is as follows. 

 

 
Figure 1. Cumulative probability distribution chart of investment prospect 

 
Observation can be seen from figure 1 investment prospect 

value is mainly distributed between 0 and 0, and the 
distribution is uniform, prospect value above 0 and below 0 
companies accounted for about 10%, thus the enterprise 
actual prospect distribution between span, according to 
establish investment level is not only focused on optimistic 
rating, it is also more in line with the actual situation of the 
stock market, such as previously found that analysts release 
investment rating will be affected by institutional 
shareholding, the higher the shareholding the more inclined 

to issue optimistic rating[15], Xu Xinyi, Leung Shangkun and 
others have also proved that underwriting relations, analysts' 
reputation and insider control will all stimulate analysts to 
issue optimistic ratings[16-17]It shows that the investment 
rating report issued by the analyst lacks a certain objectivity, 
and the prospect value calculated according to the method of 
this paper suggests that the investment rating is more 
objective from the probability distribution. 

Descriptive statistical analysis of the four quarters gives the 
following table 6 

 
Table 6. Descriptive statistical analysis factor weighting of investment prospect value 

 1806 1903 1906 1909 
mean 0.376836 0.442334 0.361647 0.371782 

variance 0.059358 0.043844 0.078234 0.061349 
least value 0.240059 0.337684 0.234784 0.260864 

median 0.368174 0.435122 0.33871 0.357693 
crest value 0.563666 0.586032 0.657812 0.749097 

Observations 352 352 352 352 



 

11 

 
As can be seen in the above table, the overall level of the 

investment outlook value of each enterprise in the first quarter 
of 2019 is higher than that of other quarters, while the overall 
level of the other three quarters is not much different, and the 
distribution of the outlook value of each quarter is around the 
average, with little fluctuation. 

4. Forecast Test of Investment Rating 
This chapter according to the previous chapter of four 

quarters of investment prospect value, each quarter according 
to the outlook value from high to low, due to the W ind 
comprehensive rating stock rating is divided into five 
categories, 13 levels, so the investment prospect value is 
divided into 13 groups to build portfolio, and select the 
portfolio asset quality inspection index for inspection. 

4.1. Selection of Predictive Test Indicators 
Reviewing the relevant literature, this paper selects 

performance indicators: interval return; performance 
evaluation indicators: annualized volatility, A lpha, B eta, 
Sharp ratio, Jensen index to test the portfolio constructed by 
13 investment levels. 

Range reporting rate: for the investment income obtained 
by investors in the investment range segment, the higher the 
general range return, the higher the return brought by the 
portfolio; 

Annualized volatility: The larger the annualized volatility 
value, the return is far from the past average, and the return is 
unstable, the higher the risk, the smaller the general 
annualized volatility, the combined return is stable and the 
less the risk. 

Alpha value: it represents non-systemic risk, namely excess 
return, which has nothing to do with market fluctuations, not 
by systemic rise. The higher the A lpha, the higher the excess 
return obtained by the investment strategy. 

Beta value: represents systemic risk, that is, the sensitivity 
of investment strategy performance to the changes of the 
market. The higher the B eta, the more volatile the stocks are 
relative to the market index. 

Sharp ratio: indicates how much excess return per unit of 
risk.If it is positive, the fund return rate is higher than the 
volatility risk; if negative, the operational risk is too return. 
The higher the proportion, the better the portfolio. 

Jensen index: For each excess return from one unit of 
systemic risk, the greater the Jensen index, the better the 
portfolio strategy performance. 

4.2. Investment Rating Inspection 
Also, for example, in forecasting 2019, Taking the 

comprehensive evaluation score obtained in the fourth quarter 
of 2018 by the entropy weight Topsis method as the 

foreground value for the first quarter of 2019, And according 
to the investment outlook value from high to low is divided 
into 13 groups, Corresponding to the 13 investment levels, 
respectively, In order, buy, buy-, increase +, increase, 
increase-, neutral +, neutral, neutral-, reduction +, reduction, 
reduction-, sell +, sell, Build 13 investment portfolios; Also 
construct the portfolio with the composite evaluation score for 
the first quarter of 2019 as the outlook value for the second 
quarter of 2019, and so on.Each investment level corresponds 
to a strategy portfolio, the portfolio holding period is a quarter, 
the last trading day of each quarter adjusts the portfolio of the 
next quarter, a total of 4 times, and then the value of each test 
index of each portfolio in 2019 is calculated, and the results 
are shown in the following figure below. 

As can be seen in the figure above analysis, the abscissa 
represents the investment level, decreasing in order from left 
to right; the ordinate represents the value of the test index of 
each portfolio.From the overall trend, During the projected 
2019 period, For the range rate of return, The return rate of 
high investment level is higher than the low investment level; 
A lpha values indicate that the portfolio is not rewarded by 
market volatility, Portfolio with high investment level have 
higher A lpha values than those with low investment level; 
The S harp e ratio is an indicator that combines both earnings 
and risk, As can be seen from the figure above, most level 
high portfolio Sharpe ratio values are greater than 1, It 
indicates that the combined return rate is high over the 
fluctuation risk; The overall trend of the Jensen index has also 
decreased with the decrease of the investment level, It 
indicates that the portfolio with high investment level obtains 
higher excess return from a unit of system risk than the 
portfolio with low investment level.Similarly, on the overall 
trend, for annualized volatility and Beta values, from high-
rated buy combinations to low-rated sell combinations, 
indicating that high-grade portfolio returns are more stable, 
less risky than low combinations and smoother volatility than 
the broader index.On the whole, the portfolio with high 
investment rating is better than that with low investment 
rating, indicating that it is feasible to select the prospect 
attribute factor according to the forward segment regression 
method, and then calculate the investment prospect value 
according to the entropy right Topsis method to predict the 
investment rating method, which can provide investment 
decision suggestions for investors. 

According to the same method steps screening the 
investment prospect attribute factor, calculate investment 
prospect value and build portfolio to forecast in 2017 and 
2018 investment rating, get the rating of each index although 
there are local fluctuations, but the overall trend and forecast 
2019 investment prospects have similar investment rating, 
high investment rating portfolio indicators better than low 
investment level portfolio.

 



 

12 

 

 

 
Figure 2. Inspection diagram of each index of investment rating 

 

5. Conclusions 
The stock market has been a complex system difficult to 

predict and has been changing.Based on our research at home 
and abroad, Through a forward-direction segmented 
regression approach, Using four quarters as the study time 
window, Fitting the factors related to the investment ratings 
for the current quarter and the W ind composite rating for the 
next quarter, Ftotal fit 4 times, Dynamic filtering out of the 
foreground attribute factor in this time window, Then the 
entropy right Top sis method is used to calculate the 
comprehensive rating score of each quarter as the investment 
prospect value of the next quarter, Then build the portfolio at 
13 levels from high to low, Finally, interval returns, 
annualized volatility, Alpha, Beta, Sharpe ratio and Jen sen 
index are selected to test the asset quality of portfolportfolio 
constructed by different investment levels. 

The experimental results show that from the overall trend, 
the portfolio strategy with high investment rating is better 
than the portfolio strategy with low investment rating, which 
can provide the investment decision suggestions such as 
buying and selling. Moreover, the factor is not fixed, but 
changes dynamically with different business conditions in the 
time window, and the forecast results are not fixed and 
consistent with the characteristics of the stock market change. 

References 
[1] Jin (2020) Financial market review in the first half of the year 

[J].China Finance,no,15,p.79-80. 

[2] Xie Lingling(2020).Retail and institutional investor differences 
and analyst reporting strategy —is empirical evidence from the 
Chinese market [J].Shanghai Finance, no.10, p.58-68. 

[3] Zhang Ran, Wang Rongfei, Wang Shenghua(2017).Analyst 
correction information, fundamental analysis, and future stock 
earnings [J].Financial Research,no.07,p.156-174. 

[4] Huang Zheng(2020).Predictability of stock earnings: factors 
from the literature [J].International Business Accounting, no.09 
p.92-96. 

[5] Banz  R W(1981).The relationship between return and market 
value of common stocks[J].Journal of Financial Economics, 
vol.9 no.1, p.3-18. 

[6] Grace Xing Hu et al(2019).Fama –French in China: Size and 
Value Factors in Chinese Stock Returns[J].International 
Review of Finance,vol.09, no.1,p.3-44. 

[7] Zhao Mengxue, Tu Suping, Yang Keyu(2020).Scale Effect and 
Value Effect —— is empirical evidence from China's A-share 
market [J].Modern Marketing (next ten days issue), no.06,p.44-
45. 



 

13 

[8] Fama  E F, French K R(1992).The Cross-Section of Expected 
Stock Returns[J].The Journal of Finance, vol.47, no.02p.427-
465. 

[9] Edirisinghe  N C P, Zhang X(2007),.Generalized DEA model 
of fundamental analysis and its application to portfolio 
optimization[J].Journal of banking & finance, vol.31, no.11, 
p.3311-3335 

[10] Liu Jianrong, Pan Heping(2010).Analysis of Investment Value 
of Chinese Listed Companies Based on Financial Index System 
[J].Administrator (Academic edition), no.05,p.16-25. 

[11] Pal Rudrajeet , Kupka  Karel, Aneja  Arun P, Militky  
Jiri(2016).Business  health characterization: A hybrid 
regression and support vector machine analysis[J].Expert 
Systems with Applications, vol.49.al 

[12] Chen Xiaoxin(2004).The Application of the Entropy Weight 
Method in the Stock Market [J].Business Research, 
no.16,p.139-140 + 144. 

[13] Zhang Sui, Zhang Mei, Chi Guotai(2010).Scientific and 
Technical Evaluation Model based on Entropy Weight Method 
[J].Journal of Management, vol.07,no.01,p.34-42. 

[14] C.L.Hwang and K.Yoon(1981).Multiple Attribute Decision 
Making[M].Spring Verlag.Berlin 

[15] Zhu Weidong, Wang Lina, Shen Jie(2016).Research on the 
optimistic tendency of securities analysts under the 
shareholding pressure of institutional investors 
[J].Management Science in China, vol.24,no.08, p.45-52. 

[16] Wang Huifang, Xu Xinyi(2017).Optimistic research of 
securities analysts under private placement [J].Journal of 
Nanjing Audit University,vol.14, no.03,p.47-56. 

[17] Sun Shuwei, Liang Shangkun, Fu Yuxiang, Zhu 
Hongjun(2019).Choose the opportunity or conspiracy —— 
insiders before the analyst behavior [J].Journal of Management 
Science, vol.22,no.12, p.105-123.