TWI708202B - Stock-to-bond ratio trend prediction system and stock-to-bond ratio trend prediction method - Google Patents

Abstract

A stock-to-bond ratio trend prediction system is adapted to predict an operation strategy of historical stock-to-bond ratio data for a next transaction date. The prediction system includes a processor, a storage module, and an output module signally connected with each other. Accordingly, the processor is adapted to build a survival probability model to estimate possible developing trend for the stock-to-bond ratio so as to provide proper operation strategy. A stock-to-bond ratio trend prediction method is also provided.

Description

System and method for forecasting stock-debt ratio trend

The present invention relates to a related technology of financial product investment prediction, and particularly relates to a system and method for predicting the trend of equity-debt ratio.

With the improvement of human living standards, in addition to earning active income through work, more and more people are earning passive income through investment and financial management. The list of eligible investment targets is endless, and various financial products (such as stocks, bonds, funds, etc.) are more popular due to the relatively abundant public information.

In addition to investing in a single financial product, some people also use asset portfolios to perform compound investment operations to improve investment performance. The asset portfolio can include inherently higher risk stocks and lower risk bonds.

In order to be able to effectively predict the rise and fall trends of asset portfolios including stocks and bonds, in one embodiment, a stock-to-debt ratio trend prediction system is suitable for predicting the operation strategy of the next trading day of multiple historical stock-to-debt ratio data. The prediction system includes a processor, a storage module, and an output module that are signal-connected to each other. The processor obtains the historical equity-debt ratio data from the storage module, and each historical equity-debt ratio data includes a plurality of characteristic values. The processor builds a survival probability model based on the historical equity-debt ratio data. The survival probability model includes a survival probability of each historical equity-debt ratio data in multiple predicted trends, and is determined by the highest survival probability among the historical equity-debt ratio data. The corresponding forecast trend is the selected forecast trend of each historical equity-debt ratio data. The processor determines the operation strategy for the next trading day based on the selected forecast trends of the historical equity-debt ratio data.

In one or more embodiments, the processor further builds a survival probability model based on the historical equity-to-debt ratio data and at least one piece of market information. Further, the processor further uses the historical equity-to-debt ratio data and at least one piece of market information as input values of the random survival forest algorithm to establish a survival probability model.

In one or more embodiments, the processor further divides the historical equity ratio data into a plurality of first historical equity ratio data and a plurality of second historical equity ratio data according to different conditions of each characteristic value , And the processor builds a survival probability model based on the first historical equity ratio data. Furthermore, the processor obtains the parent confusion matrix and multiple sub-sample confusion matrices based on the second historical equity-to-debt ratio data, where the parent confusion matrix corresponds to the selected prediction trend and has a maternal accuracy rate, and each sub-sample confusion matrix corresponds to the selected The prediction trend has a sub-sample accuracy rate. The processor also judges that each first historical equity-to-debt ratio data meets the grouping conditions. The grouping conditions correspond to the confusion matrix of the selected sub-samples in the confusion matrix of the sub-samples. The selected forecast trend of the higher of the accuracy of the parent sample and the sub-sample accuracy of the confusion matrix of the selected sub-sample is used as the selected forecast trend of each first historical equity-debt ratio data.

In one or more embodiments, when the highest survival probability in each first historical equity-debt ratio data is greater than the high threshold value, the processor determines that each first historical equity-debt ratio data meets the grouping condition. Further, the high threshold is between 0.6 and 0.9.

In one or more embodiments, when the survival probability of each of the first historical equity-debt ratio data corresponding to one of the predicted trends is greater than the low threshold value, the processor determines that each first historical equity-debt ratio data satisfies Grouping conditions. Further, the low threshold is between 0.4 and 0.5.

In one or more embodiments, the processor further performs a moving weighted average calculation on the selected forecast trends corresponding to the multiple recent equity-debt ratio data before the next trading day of the historical equity-debt ratio data. The strength value of the recent equity-debt ratio data, and the processor further determines the next trading day based on the strength value of the recent equity-debt ratio data and the standard deviation of the selected forecast trends of the recent equity-debt ratio data Operational strategy.

In another embodiment, a method for predicting the trend of stock-to-debt ratio is applicable to the operation strategy for predicting multiple historical stock-to-debt ratio data on the next trading day. The prediction method includes the step of obtaining historical stock-to-debt ratio data: obtaining the historical stocks Debt ratio data, where each historical equity-debt ratio data includes multiple characteristic values; the steps for establishing a survival probability model: use the historical equity-debt ratio data to establish a survival probability model, where the survival probability model includes each historical equity-debt ratio data. The survival probability of a predicted trend, and the predicted trend corresponding to the highest survival probability in each historical stock-debt ratio data is the selected forecast trend of each historical stock-debt ratio data; and the operation strategy decision step: use these historical stocks The selected forecast trend of the data determines the operation strategy for the next trading day.

In one or more embodiments, the step of establishing the survival probability model further includes: establishing a survival probability model based on the historical equity-debt ratio data and at least one market information. Furthermore, the step of establishing the survival probability model further includes: using the historical equity-to-debt ratio data and at least one market information as the input value of the random forest algorithm to establish the survival probability model.

In one or more embodiments, the step of establishing the survival probability model further includes: according to different conditions of each characteristic value, dividing the historical equity-debt ratio data into a plurality of first historical equity-debt ratio data and A plurality of second historical equity-debt ratio data; and establishing a survival probability model based on the first historical equity-debt ratio data. Furthermore, the step of establishing the survival probability model further includes: obtaining a parent confusion matrix and a plurality of sub-sample confusion matrices based on the second historical equity-to-debt ratio data, wherein the parent confusion matrix has a maternal accuracy rate corresponding to the selected predicted trend , Each sub-sample confusion matrix has a sub-sample accuracy rate corresponding to the selected forecast trend; judging that each first historical equity-debt ratio data meets the grouping condition, where the grouping condition corresponds to the selected sub-sample confusion matrix in some sub-sample confusion matrix; and The selected prediction trend of the higher of the accuracy of the parent sample of the parent confusion matrix and the sub-sample accuracy of the selected sub-sample confusion matrix is used as the selected prediction trend of each first historical equity-debt ratio data.

In one or more embodiments, the step of establishing the survival probability model further includes: when the highest survival probability in each first historical equity ratio data is greater than a high threshold, judging each first historical equity ratio The data meets the grouping conditions. Furthermore, the high threshold is between 0.6 and 0.9.

In one or more embodiments, the step of establishing the survival probability model further includes: when the survival probability of one of the predicted trends corresponding to each first historical equity ratio data is greater than the low threshold, It is judged that the data of the first historical equity-debt ratio meets the grouping conditions. Furthermore, the low threshold is between 0.4 and 0.5.

In one or more embodiments, the step of determining the operation strategy further includes: the selected forecasts corresponding to the multiple recent equity-debt ratio data before the next trading day of the historical equity-debt ratio data The trend is calculated by moving weighted average to obtain the intensity value of the recent equity-debt ratio data; and determined based on the intensity value of the recent equity-debt ratio data and the standard deviation of the selected forecast trends of the recent equity-debt ratio data Operation strategy for the next trading day.

In summary, the system and method for predicting the trend of equity-debt ratio according to one or more embodiments of the present invention use a survival probability model to estimate possible trends and provide appropriate operating strategies. In some embodiments, the historical stock-to-debt ratio data can be divided into the first historical stock-to-debt ratio data and the second historical stock-to-debt ratio data, and the first historical stock-to-debt ratio data is used as training data to input the model, and the 2. The historical equity-to-debt ratio data is used as the verification data to generate the parent confusion matrix and the sub-sample confusion matrix, and the first historical equity-to-debt ratio data is contextually cut through the grouping conditions to improve the accuracy of prediction. Furthermore, in one or more embodiments, the output signal can be made smoother by means of moving weighted average, so there is no need to frequently change the ratio of stocks and debts, which can effectively save transaction costs and increase profits.

As shown in FIG. 1, it shows a stock-to-debt ratio trend prediction system (hereinafter referred to as the forecasting system 80) according to an embodiment of the present invention, which is suitable for an operation strategy for predicting multiple historical stock-to-debt ratio data on the next trading day. In other words, the forecasting system 80 can provide an operating strategy for the next trading day, which can be active, stable, or conservative, and the user can then adjust the allocation of the ratio of stocks to bonds held according to the provided operating strategy. The stock-to-debt ratio information includes the transaction date and the stock-to-debt ratio. The stock-to-debt ratio is the divided value of the return on stocks and the return on bonds. When the stock-to-debt ratio trend is rising, it means that the stock ratio should continue to be increased while reducing the bond ratio for active operations; on the other hand, when the stock-to-debt ratio trend is declining, it means that the bond ratio should be increased while reducing the stock ratio for conservative operations.

As shown in FIG. 1, the prediction system 80 includes a processor 81, a storage module 82 and an output module 83. The processor 81, the storage module 82, and the output module 83 are connected to each other through wired or wireless signals. For example, the prediction system 80 may be an industrial computer, a personal computer, a notebook computer, a smart phone, a tablet computer, etc. Here, the processor 81 may be implemented by one or more processing elements. Here, each processing element can be a microprocessor, a microcontroller, a digital signal processor, a microcomputer, a central processing unit, a field programming gate array, a programmable logic device, a state machine, a logic circuit, an analog circuit, a digital circuit, and /Or any device based on the operation command signal (analog and/or digital), but it is not limited here. The storage module 82 may be realized by one or more storage elements. Among them, each storage element can be, for example, a non-volatile memory, a hard disk, an optical disk, or a tape, but it is not limited here. In addition, it should be noted that the data stored in the storage module 82 can come from a terminal connected via a wired or wireless connection, and can be updated in real time or in batches. The output module 83 can be a screen, a printer, or a voice output device (such as a speaker), but it is not limited here. The following is a further description of the functions of each component in this prediction system.

First, the processor 81 obtains historical equity-debt ratio data from the storage module 82. The stock-to-debt ratio data can be used to reflect the stock-to-debt ratio of a certain investment portfolio. Among them, each historical stock-to-debt ratio data includes multiple characteristic values. The characteristic value includes the stock holding ratio of this portfolio and the corresponding technical indicators (such as but not limited to MACD, KD, Bollinger Bands, RSI, etc.). The higher the stock holding ratio reflects the lower the bond holding ratio, and the more active investment strategy; the lower the stock holding ratio reflects the higher the bond holding ratio, the more conservative the investment strategy; if it is stocks and bonds If the holding ratio is equal, the investment strategy is relatively stable.

Then, the processor 81 establishes a survival probability model based on the historical equity-debt ratio data. The survival probability model includes the survival probability of each historical stock-to-debt ratio in an upward trend, a downward trend, and a consolidation trend. The predicted trend corresponding to the highest survival probability in each historical equity-debt ratio data is the selected forecast trend of each historical equity-debt ratio data. For example, take Table 1 below as an example. Table 1 shows some of the historical stock-to-debt ratio data obtained after the survival probability model, including the historical stock-to-debt ratio data are rising, falling and consolidating. The survival probability of the trend, and the trend corresponding to the highest survival probability is the selected predicted trend. Taking the data on November 1, 2018 as an example, the survival probability of the stock-to-debt ratio data on the day in a downward trend is 0.18666667, the survival probability in a consolidation trend is 0.34666667, and the survival probability in an upward trend is 0.46666667. Therefore, the selected forecast trend of the stock-to-debt ratio data on November 1, 2018 is upward, and the rest of the day will not be repeated by analogy.

[Table 1] transaction date Selected forecast trend Probability of falling Consolidation probability Probability of rising 20181101 1 0.18666667 0.34666667 0.46666667 20181102 0 0.21333333 0.41666667 0.37000000 20181105 1 0.28000000 0.35666667 0.36333333 20181106 1 0.21333333 0.28000000 0.50666667 20181107 1 0.20333333 0.18000000 0.61666667 20181108 1 0.20666667 0.25333333 0.54000000 20181109 0 0.22333333 0.45000000 0.32666667 20181112 0 0.31666667 0.45666667 0.22666667 20181113 0 0.35666667 0.40666667 0.23666667 20181114 -1 0.55000000 0.19333333 0.25666667

Finally, the processor 81 determines the operation strategy for the next trading day based on the selected predicted trends of the historical equity-debt ratio data, and the output module 83 outputs the operation strategy. For example, assuming that the last piece of historical equity-to-debt ratio data is November 14, 2018, and the selected forecast trend on that day is down, the processor 81 can decide based on the selected forecast trend on November 14, 2018 The operation strategy of the next trading day (for example, November 15, 2018) is output through the output module 83. Specifically, when the output module 83 is a display screen, the operation strategy can be presented with text, image, video, or a combination of the above, but it should be understood that it is not limited to this; as mentioned above, the output module 83 can also be a printer The operation strategy can be printed out by the machine, or the output module 83 can also be a speaker to present the operation strategy in the form of sound or music. On the other hand, in this example, the processor 81 can also determine the operation on November 15, 2018 based on the average of the selected forecast trends on November 14, 2018 and the days before November 14, 2018 Strategy.

In one or more embodiments of the present invention, the processor 81 further establishes a survival probability model based on the historical equity-to-debt ratio data and at least one piece of market information. For example, market information can be risk sentiment, such as interest rate spreads, exchange rates, VIX index, etc.; market information can also be various general economic indicators, such as stock price index, price-to-earnings ratio, yield rate, consumer price index, etc. Furthermore, in one or more embodiments, the processor 81 further uses the historical equity-to-debt ratio data and the aforementioned market information as input values of the random survival forest algorithm to establish a survival probability model. It should be noted that although the random survival forest algorithm is used to establish the survival probability model here, the present invention is not limited to this algorithm, as long as the survival probability can be predicted.

In addition, because the historical stock-to-debt ratio data has actual trends (for example, on November 2, 2018, the selected forecast trend is consolidation, while the actual trend is upward). Therefore, the type and/or weight of the input value of the adjustment algorithm can be used to verify the algorithm, so that the generated survival probability model can better reflect the actual situation and provide users with more accurate prediction analysis.

In addition, in one or more embodiments of the present invention, the processor 81 further divides the historical equity-debt ratio data into a plurality of first historical equity-debt ratio data and multiple data according to different conditions of each characteristic value. The second historical stock-to-debt ratio information. For example, in one embodiment, the processor 81 divides the historical equity ratio data into a plurality of first historical equity debts based on the condition that the transaction date of the historical equity debt ratio data is earlier than March 25, 2015 Ratio data (the stock-to-debt ratio data with the transaction date earlier than March 25, 2015) and multiple second historical stock-to-debt ratio data (the stock debt ratio with the transaction date later than March 25, 2015), and the processor 81 also established a survival probability model based on the first historical equity-debt ratio data. It should be noted that although in this embodiment, the processor 81 divides the historical equity ratio data into two parts based on the transaction date as a condition, it is not limited to this; in some embodiments, the processing The device 81 may also obtain multiple sample groups by sampling historical equity-debt ratio data, and then divide the sample groups into the first historical equity-debt ratio data and the second historical equity-debt ratio data.

Then, the processor 81 further obtains a parent confusion matrix and a plurality of sub-sample confusion matrices based on the second historical equity-to-debt ratio data. The parent confusion matrix corresponds to the selected prediction trend and has a parent accuracy rate, and each sub-sample confusion matrix corresponds to the selected prediction The trend has a sub-sample accuracy rate. In this embodiment, the matrix confusion matrix is the confusion matrix obtained based on all the second historical equity-to-debt ratio data, which includes model forecast trends corresponding to rising, consolidation, and falling, actual trends, and the average number of samples corresponding to various situations. The sub-sample confusion matrix is a confusion matrix obtained by selecting a part of all the second historical equity-debt ratio data through different grouping conditions. It also includes model forecast trends corresponding to rising, consolidation, and falling, actual trends, and corresponding situations. Average number of samples. Specifically, the aforementioned grouping condition is that when the processor 81 determines that the highest probability of each predicted trend in a second historical equity-to-debt ratio data is greater than the high threshold (for example, between 0.6 and 0.9), the processor 81 applies the second The historical equity-to-debt ratio data is grouped into sub-confusion matrices Q1 to Q4 to correspond to different scenarios. Furthermore, when the highest probability is greater than 0.9, group to the Q1 sub-confusion matrix; when the highest probability is greater than 0.8, group to the Q2 sub-confusion matrix; when the highest probability is greater than 0.7, group to the Q3 sub-confusion matrix; when the highest probability is greater than At 0.6, group into Q4 sub-confusion matrix. On the other hand, when the processor 81 determines that the probability of the selected predicted trend in the second historical equity ratio data is greater than the low threshold (for example, between 0.4 and 0.5), the processor will use this second historical equity The data is grouped into sub-confusion matrices from Q5 to Q11 to correspond to different situations. Furthermore, when the probability of the selected forecast trend rising is greater than 0.5, group it into the Q5 sub-confusion matrix; when the probability of the selected forecast trend being consolidation is greater than 0.5, group it into the Q6 sub-confusion matrix; when the selected forecast trend is When the probability of falling is greater than 0.5, it is grouped into the Q7 sub-confusion matrix; when the probability of the selected forecast trend rising is greater than 0.4, it is grouped to the Q8 sub-confusion matrix; when the probability of the selected forecast trend being consolidation is greater than 0.4, grouped to Q9 Sub-confusion matrix; when the probability of the selected predicted trend falling is greater than 0.4, group it into the Q10 sub-confusion matrix, otherwise group it into the Q11 sub-confusion matrix.

Then, the processor 81 then determines which of the grouping conditions each first historical equity-debt ratio data meets to determine the selected sub-sample confusion matrix corresponding to each first historical equity-debt ratio data, and obtains each first in Table 2 below. Historical stock-to-debt ratio data and corresponding scenarios. In other words, the processor 81 uses the aforementioned grouping conditions to determine the situation to which each first historical equity ratio data should be classified, which will not be repeated here.

[Table 2] transaction date Selected forecast trend Probability of falling Consolidation probability Probability of rising Situation 20181101 1 0.18666667 0.34666667 0.46666667 Q10 20181102 0 0.21333333 0.41666667 0.37000000 Q9 20181105 1 0.28000000 0.35666667 0.36333333 Q11 20181106 1 0.21333333 0.28000000 0.50666667 Q7 20181107 1 0.20333333 0.18000000 0.61666667 Q4 20181108 1 0.20666667 0.25333333 0.54000000 Q7 20181109 0 0.22333333 0.45000000 0.32666667 Q9 20181112 0 0.31666667 0.45666667 0.22666667 Q9 20181113 0 0.35666667 0.40666667 0.23666667 Q9 20181114 -1 0.55000000 0.19333333 0.25666667 Q5

After determining the context of each first historical equity-debt ratio data (ie, the confusion matrix of the selected sub-sample), the processor 81 then compares the accuracy of the parent sample of the parent confusion matrix with the accuracy of the sub-sample of the selected sub-sample confusion matrix The selected forecast trend of the high person is used as the selected forecast trend of the first historical stock-to-debt ratio data, as shown in Table 3 and Table 4 below.

Table 3 shows the matrix confusion matrix based on a total of 2,200 stock-to-debt ratio data from June 27, 2006 to March 24, 2015. As shown in Table 3, the average number of samples whose predicted trend is down (represented by -1) and the actual trend is down (ie probability) is 42.57%, and the average sample whose predicted trend is down and the actual trend is consolidation (represented by 0) The number is 30.69%, and the average number of samples whose predicted trend is down and the actual trend is up (represented by 1) is 24.09%. Tables 4 to 14 are the confusion matrices of 11 sub-samples from Q1 to Q11 based on a total of 937 stock-to-debt ratio data from March 25, 2015 to November 30, 2018. As shown in Table 12, for example, the sub-sample confusion matrix Q9 has a total of 185 samples, of which the average number of samples whose predicted trend is consolidation and the actual trend is consolidation is 42.26%, and the predicted trend is upward and the actual trend is upward. The average sample size is 35.29%. Therefore, for the stock-to-debt ratio data on November 7, 2018, the accuracy of the parent confusion matrix is 50.14%, which is 42.26% greater than the accuracy of the sub-sample confusion matrix Q9. In other words, the probability is that on the premise that the predicted trend is down, the actual decline is 42.57%, the actual consolidation is 30.69%, and the actual increase is 24.09% (actually a conditional probability, not a simple probability).

[table 3] Maternal confusion matrix actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 42.57 30.69 24.09 Consolidation (0) 18.84 50.14 31.01 Up (1) 12.94 37.06 49.65

[Table 4] Q1 n=1 actual Consolidation (0) prediction Down (-1) 100

[table 5] Q2 n=23 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 58.82 23.53 17.65 Consolidation (0) 100.00 0.00 0.00 Up (1) 0.00 60.00 40.00

[Table 6] Q3 n=55 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 71.43 14.29 14.29 Consolidation (0) 37.50 62.50 0.00 Up (1) 5.26 52.63 42.11

[Table 7] Q4 n=153 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 36.96 36.96 26.09 Consolidation (0) 15.91 61.36 22.73 Up (1) 12.70 25.40 61.90

[Table 8] Q5 n=71 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 46.48 36.62 16.90

[Table 9] Q6 n=95 actual Down (-1) Consolidation (0) Up (1) prediction Consolidation (0) 13.68 57.89 28.42

[Table 10] Q7 n=83 actual Down (-1) Consolidation (0) Up (1) prediction Up (1) 3.61 33.73 62.65

[Table 11] Q8 n=95 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 38.71 31.18 30.11 Consolidation (0) 0.00 100.00 0.00 Up (1) 100.00 0.00 0.00

[Table 12] Q9 n=195 actual Down (-1) Consolidation (0) Up (1) prediction Consolidation (0) 22.02 42.26 35.71 Up (1) 11.76 52.94 35.29

[Table 13] Q10 n=71 actual Down (-1) Consolidation (0) Up (1) prediction Up (1) 22.54 40.85 36.62

[Table 14] Q11 n=93 actual Down (-1) Consolidation (0) Up (1) prediction Down (-1) 34.21 31.58 34.21 Consolidation (0) 13.79 48.28 37.93 Up (1) 23.08 42.31 34.62

In each sub-sample confusion matrix, the prediction result can be further adjusted according to the probability distribution of the confusion matrix. For example, if the preliminary forecast result is an increase and the scenario is Q9, since the forecast is an increase in the Q9 scenario but the actual trend has the highest probability of consolidation (52.94%), the prediction result is changed from an increase to a consolidation. Conversely, if the preliminary forecast result is a decline and the scenario is Q3, since the forecast is a fall in the Q3 scenario and the actual trend still has the highest probability of a fall (71.43%), the predicted result remains a fall.

Taking the stock-to-debt ratio data on November 7, 2018 as an example, the selected forecast trend is down and belongs to scenario Q4, then the maternal accuracy rate of the maternal confusion matrix corresponding to the downward trend is 42.57%, and the sub-sample confusion matrix Q4 corresponds to The accuracy rate of the downtrend sub-sample is 36.96%. Therefore, the processor selects the selected forecast trend of the matrix confusion matrix as the selected forecast trend of the stock-to-debt ratio data on November 7, 2018, that is, the downward trend.

In this way, the prediction system according to one or more embodiments of the present invention can verify and modify the prediction results of the model by means of context analysis, thereby better reflecting actual conditions and providing users with more accurate prediction analysis.

In one or more embodiments, the processor further performs a moving weighted average calculation on the selected forecast trends corresponding to the multiple recent equity-debt ratio data before the next trading day of the historical equity-debt ratio data. The strength value of these recent equity-debt ratio data. For example, assuming that the next trading day is November 15, 2018, the processor can perform a moving weighted average based on the selected forecast trend corresponding to the stock-to-debt ratio data 20 days before this date to obtain stocks and bonds for the past 20 days The change status of the selected prediction trend (ie intensity value) of the ratio data. When the intensity value is less than 1 and greater than or equal to 0.4, the defined intensity value is in the first interval; when the intensity value is less than 0.4 and greater than or equal to 0, the defined intensity value is in The second interval; when the intensity value is less than 0 and greater than or equal to -0.4, the defined intensity value is in the third interval; and when the intensity value is less than -0.4 and greater than or equal to -1, the defined intensity value is in the fourth interval.

Then, the processor 81 further determines the standard deviation of the recent equity-debt ratio data for the selected predicted trends. When the standard deviation of the selected forecast trend is greater than 0 but less than or equal to 1, it is considered low volatility, which means that the selected forecast trend has small changes in the recent equity-debt ratio data; on the other hand, when the selected forecast When the standard deviation of the trend is greater than 1, it is considered high volatility, which means that the selected forecast trend has a large change in the recent stock-to-debt ratio data.

Then, the processor 81 determines the operation strategy for the next trading day based on the intensity value of the recent equity-debt ratio data and the standard deviation of the selected predicted trends of the recent equity-debt ratio data. For example, when the strength value is in the first interval or the second interval with high volatility or in the first interval with low volatility, the processor 81 determines that the operation strategy of the next trading day is positive, and provides stock bonds Operational recommendations with a ratio configuration of 7:3; when the strength value is in the second or third range and has low volatility, the processor 81 determines that the operation strategy for the next trading day is robust, and provides a stock-to-debt ratio configuration of 5 : 5 operation suggestions; and when the strength value is in the third or fourth interval with high volatility or in the fourth interval with low volatility, the processor 81 decides that the operation strategy of the next trading day is conservative, and Provide operational suggestions with a debt ratio of 3:7. In this way, the output signal can be made smoother through the moving weighted average method, which means that investors do not need to frequently change the ratio of stocks and debts, and can effectively save transaction costs and increase profits.

Referring to FIG. 2, FIG. 2 is a flow chart of the steps of the method for predicting the trend of stock-to-debt ratio according to the second embodiment of the present invention, which is suitable for predicting the operation strategy of the next trading day of multiple historical stock-to-debt ratio data. As shown in FIG. 2, one or more embodiments of the present invention further disclose a method for predicting the price trend of financial commodities, including the following steps: obtaining historical equity ratio data S101, establishing survival probability model S102, and determining operation strategy S103 , The steps are explained as follows.

The step of obtaining historical stock-to-debt ratio data includes obtaining the historical stock-to-debt ratio data, as shown in step S101; wherein each historical stock-to-debt ratio data includes multiple characteristic values. The description of the eigenvalues has been described above, and will not be repeated.

The step of establishing a survival probability model includes establishing a survival probability model based on the historical equity-debt ratio data, as shown in step S102; wherein the survival probability model includes the survival probability of each historical equity-debt ratio data in a plurality of predicted trends, and the survival probability is in each history The forecast trend corresponding to the highest survival probability in the stock-to-debt ratio data is the selected forecast trend of each historical stock-to-debt ratio data.

The operation strategy determination step includes determining the operation strategy for the next trading day based on the selected predicted trends of the historical equity-to-debt ratio data, as shown in step S103.

Please refer to FIG. 3, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the third embodiment of the present invention. As shown in FIG. 3, according to one or more embodiments of the present invention, the survival probability model establishment step S102 further includes: Step S1021: Establish a survival probability model based on the historical equity-debt ratio data and at least one market information.

Please refer to FIG. 4, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the fourth embodiment of the present invention. As shown in FIG. 4, according to one or more embodiments of the present invention, in the survival probability model establishment step S102, the step S1022 is further included: using the historical equity ratio data and at least one market information as input to the random forest algorithm Value and establish a survival probability model.

Please refer to FIG. 5, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the fifth embodiment of the present invention. As shown in FIG. 5, according to one or more embodiments of the present invention, in the survival probability model establishment step S102, the step S1023 is further included: according to different conditions of each characteristic value, the historical equity and debt ratio data are divided into more A first historical stock-to-debt ratio data and a plurality of second historical stock-to-debt ratio data; and step S1024: establishing a survival probability model based on the first historical stock-to-debt ratio data.

Furthermore, please refer to FIG. 6, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the sixth embodiment of the present invention. As shown in FIG. 6, according to one or more embodiments of the present invention, in the survival probability model establishment step S102, in addition to the aforementioned steps S1023 and S1024, it further includes a step S1025: obtaining the parent body based on the second historical equity-debt ratio data Confusion matrix and multiple sub-sample confusion matrices, where the parent confusion matrix has a maternal accuracy rate corresponding to the selected prediction trend, and each sub-sample confusion matrix has a sub-sample accuracy rate corresponding to the selected prediction trend; Step S1026: Determine each first historical equity ratio The data meets the grouping condition, where the grouping condition corresponds to the selected sub-sample confusion matrix in the confusion matrix of some sub-samples; and Step S1027: According to the accuracy rate of the mother sample of the parent confusion matrix and the sub-sample accuracy of the selected sub-sample confusion matrix The higher one’s selected forecast trend is used as the selected forecast trend of each first historical equity-debt ratio data.

Please refer to FIG. 7, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the seventh embodiment of the present invention. As shown in FIG. 7, according to one or more embodiments of the present invention, in step S1026, step S1028 is further included: when the highest survival probability in each first historical equity ratio data is greater than the high threshold, determine each first The historical equity-to-debt ratio information satisfies the grouping conditions. Furthermore, in one or more embodiments, the high threshold is between 0.6 and 0.9.

Please refer to FIG. 8, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the eighth embodiment of the present invention. As shown in FIG. 8, according to one or more embodiments of the present invention, in step S1026, step S1029 is further included: when each of the first historical equity-debt ratio data corresponds to one of the predicted trends, the survival probability is greater than When the threshold is low, it is judged that the first historical equity-debt ratio data meets the grouping conditions. Furthermore, in one or more embodiments, the low threshold is between 0.4 and 0.5.

Please refer to FIG. 9, which is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the ninth embodiment of the present invention. As shown in FIG. 9, according to one or more embodiments of the present invention, in the operation strategy determination step S103, the step S1031 is further included: a plurality of recent equity debt ratios before the next trading day of the historical equity debt ratio data The selected forecast trends corresponding to the data are calculated by moving weighted average to obtain the strength value of the recent equity-debt ratio data; and step S1032: According to the strength value of the recent equity-debt ratio data and the recent equity-debt ratio The standard deviation of the selected forecast trends of the data determines the operation strategy for the next trading day.

In summary, the system and method for predicting the trend of equity-debt ratio according to one or more embodiments of the present invention use a survival probability model to estimate possible trends and provide appropriate operating strategies. In some embodiments, the historical stock-to-debt ratio data can be divided into the first historical stock-to-debt ratio data and the second historical stock-to-debt ratio data, and the first historical stock-to-debt ratio data is used as training data to input the model, and the 2. The historical equity-to-debt ratio data is used as the verification data to generate the parent confusion matrix and the sub-sample confusion matrix, and the first historical equity-to-debt ratio data is contextually cut through the grouping conditions to improve the accuracy of prediction. Furthermore, in one or more embodiments, the output signal can be made smoother by means of moving weighted average, so there is no need to frequently change the ratio of stocks and debts, which can effectively save transaction costs and increase profits.

80: prediction system 81: processor 82: storage module 83: output module S101~S103, S1021~S1029, S1031~S1032: steps

[Figure 1] is a schematic diagram of the structure of the stock-to-debt ratio trend prediction system according to the first embodiment of the present invention. [Fig. 2] is a flowchart of the steps of the method for predicting the trend of equity-debt ratio according to the second embodiment of the present invention. [Fig. 3] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the third embodiment of the present invention. [Fig. 4] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the fourth embodiment of the present invention. [Fig. 5] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the fifth embodiment of the present invention. [Fig. 6] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the sixth embodiment of the present invention. [Fig. 7] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the seventh embodiment of the present invention. [Fig. 8] is a detailed flowchart of the method for predicting the trend of equity-debt ratio according to the eighth embodiment of the present invention. [Fig. 9] is a detailed flow chart of the method for predicting the trend of equity-debt ratio according to the ninth embodiment of the present invention.

80: prediction system

81: processor

82: storage module

83: output module

Claims (16)

A trend prediction system for stock-to-debt ratio, which is suitable for predicting an operation strategy for the next trading day of multiple historical stock-to-debt ratio data. The prediction system includes: a processor, a storage module, and an output module. The processor The storage module and the output module are signal-connected to each other, wherein: the processor obtains the historical equity-debt ratio data from the storage module, wherein each historical equity-debt ratio data includes a plurality of characteristic values; the processor Establish a survival probability model based on the historical equity-debt ratio data, wherein the survival probability model includes a survival probability of each historical equity-debt ratio data in a plurality of predicted trends, and the highest one among the historical equity-debt ratio data The predicted trend corresponding to the survival probability is a selected predicted trend of each historical equity-debt ratio data; the processor uses the selected predicted trends of the historical equity-debt ratio data to determine the operation on the next trading day Strategy, and the output module outputs the operation strategy; wherein the processor further divides the historical equity-debt ratio data into multiple first historical equity-debt ratio data and multiple data based on different conditions of each characteristic value The second historical equity-debt ratio data; the processor establishes the survival probability model based on the first historical equity-debt ratio data; wherein the processor further obtains a matrix confusion matrix and the number A sub-sample confusion matrix, the parent confusion matrix has a maternal accuracy rate for the selected prediction trend, and each sub-sample confusion matrix has a sub-sample accuracy rate for the selected prediction trend; the processor judges each of the first historical stocks The debt ratio data satisfies a grouping condition, and the grouping condition corresponds to the confusion matrix of one of the sub-sample confusion matrices; the processor further mixes the mother sample with the selected sub-sample according to the accuracy of the parent confusion matrix The subsample of the matrix The selected forecast trend of the higher accuracy rate is used as the selected forecast trend of each of the first historical equity-debt ratio data. The prediction system according to claim 1, wherein the processor establishes the survival probability model based on the historical equity-debt ratio data and at least one market information. The prediction system according to claim 2, wherein the processor uses the historical stock-to-debt ratio data and the at least one market information as input values of the random survival forest algorithm to establish the survival probability model. The prediction system according to claim 1, wherein when the highest survival probability among the first historical equity-debt ratio data is greater than a high threshold value, the processor determines that each of the first historical equity-debt ratio data satisfies the grouping condition. The prediction system according to claim 4, wherein the high threshold is between 0.6 and 0.9. The prediction system according to claim 1, wherein when the survival probability of each of the first historical equity-debt ratio data corresponding to one of the predicted trends is greater than a low threshold value, the processor determines each of the first A historical stock-to-debt ratio data satisfies the grouping condition. The prediction system according to claim 6, wherein the low threshold is between 0.4 and 0.5. The prediction system according to claim 1, wherein the processor further performs moving weighting on the selected forecast trends corresponding to the multiple recent equity-debt ratio data before the next trading day of the historical equity-debt ratio data Calculate on average to obtain a strength value of the recent equity-debt ratio data; the processor is based on the strength value of the recent equity-debt ratio data and the recent equity-debt ratio data A standard deviation of the selected predicted trends from the data determines the operation strategy for the next trading day. A method for predicting the trend of stock-to-debt ratio, which is suitable for an operation strategy for predicting multiple historical stock-to-debt ratio data on the next trading day. The forecasting method includes: the step of obtaining historical stock-to-debt ratio data: using a processor to self-save a storage module Obtain the historical equity-debt ratio data, where each historical equity-debt ratio data includes multiple characteristic values; the survival probability model establishment step: use the processor to establish a survival probability model with the historical equity-debt ratio data, wherein the survival The probability model includes a survival probability of each historical equity-debt ratio data in multiple predicted trends, and the predicted trend corresponding to the highest survival probability in each historical equity-debt ratio data is each historical equity-debt ratio data A selected predicted trend of the; and the operation strategy decision step: using the processor to determine the operation strategy of the next trading day based on the selected predicted trends of the historical equity-to-debt ratio data; wherein a model is built on the survival probability In the step, it further includes: using the processor to divide the historical equity-debt ratio data into a plurality of first historical equity-debt ratio data and a plurality of second historical equity-debt ratio data according to different conditions of each characteristic value; According to the first historical equity-to-debt ratio data, the processor is used to establish the survival probability model; based on the second historical equity-to-debt ratio data, the processor is used to obtain a matrix confusion matrix and a plurality of sub-sample confusion matrices, where the The maternal confusion matrix has a maternal accuracy rate for the selected prediction trend, and each sub-sample confusion matrix has a sub-sample accuracy rate for the selected prediction trend; Use the processor to determine that each of the first historical equity-to-debt ratio data satisfies a grouping condition, wherein the grouping condition corresponds to one of the sub-sample confusion matrices; the selected sub-sample confusion matrix; and the processor is used according to the parent confusion matrix The selected predicted trend of the higher of the accuracy of the parent sample and the sub-sample of the confusion matrix of the selected sub-sample is used as the selected predicted trend of each of the first historical equity-debt ratio data. The prediction method according to claim 9, wherein the step of establishing the survival probability model further includes: using the processor to establish the survival probability model based on the historical equity ratio data and at least one market information. The prediction method according to claim 10, wherein the step of establishing the survival probability model further includes: using the historical equity ratio data and the at least one market information as the input value of the random forest algorithm, and using the processing The device establishes the survival probability model. The prediction method according to claim 9, wherein the step of establishing the survival probability model further includes: when the highest survival probability in each of the first historical equity-to-debt ratio data is greater than a high threshold, using the processor Determine that each of the first historical stock-to-debt ratio data meets the grouping condition. The prediction method according to claim 12, wherein the high threshold is between 0.6 and 0.9. The prediction method according to claim 9, wherein the step of establishing the survival probability model further includes: When the survival probability of each of the first historical equity-debt ratio data corresponding to one of the predicted trends is greater than a low threshold value, the processor is used to determine that each of the first historical equity-debt ratio data satisfies the sub-group condition. The prediction method according to claim 14, wherein the low threshold is between 0.4 and 0.5. The forecasting method according to claim 9, wherein in the step of determining the operation strategy, it further includes: using the processor to obtain a plurality of recent stock-to-debt ratio data before the next trading day of the historical stock-to-debt ratio data. The corresponding selected forecast trends are calculated by moving weighted average to obtain a strength value of the recent equity-debt ratio data; and the strength value based on the recent equity-debt ratio data and the recent equity-debt ratio data A standard deviation of the selected predicted trends, using the processor to determine the operation strategy for the next trading day.

Images