Northbound fund NFMACD index establishment method and device with advanced feature

Abstract

The invention discloses a northbound fund NFMACD index establishment method and device with an advanced characteristic, and relates to the technical field of stock market data processing.The northbound fund NFMACD index establishment method comprises the steps that firstly, daily northbound fund net inflow data is utilized, and all technical factor indexes are calculated; then, combining the index data into a multivariate time sequence data set; thirdly, a deep learning model is constructed and trained, a data set is used as input, and net inflow data of the next day, namely the (t + 1) day, is used as a prediction target; and finally, in combination with a model prediction result and historical data, constructing an advanced NFMACD index according to a calculation method of the MACD index. The index integrates the advantages of the traditional MACD, can more effectively predict the market trend compared with the mode of purely depending on northbound fund flow data, and is enhanced in the aspects of return rate and stability; and the hysteresis problem of the traditional MACD index is solved, and the efficiency and accuracy of investment decision making are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of stock market data processing, and in particular to a method and device for establishing a northbound fund NFMACD indicator with advanced characteristics. Background Art

[0002] In recent years, the pace of opening up of China's financial market has gradually accelerated, and the depth and breadth of foreign investment in the A-share market has also expanded accordingly; as an important force of foreign investment in the A-share market, the flow of northbound funds cannot be ignored; channels such as the Shanghai-Hong Kong Stock Connect and the Shenzhen-Hong Kong Stock Connect have enabled northbound funds to flow smoothly into the Chinese mainland market, injecting new vitality into the A-share market; the inflow of northbound funds has not only brought a large amount of incremental funds to the A-share market and improved market liquidity, but more importantly, they have brought new investment concepts and strategies; these foreign institutions usually have strong financial strength, advanced investment concepts, rich investment experience and professional investment teams; when making investment decisions, they often select high-quality companies with growth potential and investment value based on in-depth fundamental analysis, rigorous valuation systems and long-term investment perspectives; therefore, the flow and holdings of northbound funds are often regarded as international investors' "votes" on China's economic prospects and corporate profitability, which have a profound impact on the investment style and trends of the A-share market.

[0003] It is generally difficult to judge market trends using only northbound capital flows data; therefore, it is particularly important to construct indicators based on northbound capital flows in combination with commonly used technical indicators in the stock market, such as MACD; the MACD indicator, proposed by Appel in 1979, reflects market dynamics by comparing the differences between long-term and short-term moving averages, and is mainly used to guide stock buying and selling; consider constructing a MACD indicator based on northbound capital flows to more accurately judge market trends.

[0004] However, the traditional MACD indicator is easy to understand and simple to calculate, but it has a large lag, which makes the accuracy of its judgment of market trends low. Summary of the invention

[0005] The embodiment of the present invention provides a method and device for establishing a NFMACD indicator for northbound funds with advanced characteristics, which can solve the problem in the prior art that the traditional MACD indicator is easy to understand and simple to calculate, but has a large lag, resulting in low accuracy in the market trend it judges.

[0006] The embodiment of the present invention provides a method for establishing a NFMACD indicator of northbound funds with a forward-looking characteristic, comprising the following steps:

[0007] Use the daily northbound fund closing net inflow data to obtain the technical factor indicator data of the net inflow data, and use the create function to splice the technical factor indicator data into a multivariate time series data set;

[0008] Construct a deep learning network model, take the daily multivariate time series data set as input, and take the net inflow data of northbound funds at the closing of the next day as the target, train the deep learning network model and obtain a prediction model;

[0009] Predict the net inflow data of northbound funds at closing time through the prediction model to obtain the predicted closing value; combine the predicted closing value with the historical net inflow closing value to reconstruct the daily net inflow data of northbound funds at closing time;

[0010] Based on the reconstructed daily northbound fund net inflow data at closing, the construction method of the MACD indicator is simulated to construct the northbound fund NFMACD indicator with advanced characteristics.

[0011] Preferably, the technical factor indicator data includes:

[0012] Long-short index, moving average difference, simple moving average, exponential average, exponential smoothing difference, momentum index, price oscillator index, deviation rate, momentum index, difference difference momentum index, rate of change index, difference difference change rate index, SRMIMI modified index and standard deviation.

[0013] Preferably, after acquiring the multivariate time series data set, a principal component analysis method is used to perform dimensionality reduction processing on the multivariate time series data set.

[0014] Preferably, the deep learning network model includes a fully connected input layer, an LSTM recurrent network and a fully connected output layer;

[0015] The number of neurons in the fully connected input layer is 4;

[0016] The number of units in the LSTM recurrent network is 4;

[0017] The weight optimizer of the deep learning network model is an SGD optimizer;

[0018] The loss function of the deep learning network model is RMSE.

[0019] Preferably, the reconstructed daily northbound fund closing net inflow data is:

[0020] new Northbound Funds(t,z)=[S(tz),···,S(t-1),S(t),P(t+1)]

[0021] Among them: t represents the date of the day; S(t) represents the net inflow of northbound funds on day t; P(t+1) represents the forecast data on day t; z represents the exponential smoothing period.

[0022] Preferably, the construction of the NFMACD indicator includes:

[0023] Based on the reconstructed daily northbound fund closing net inflow data, the slow exponential moving average EMA and fast exponential moving average EMA of the northbound fund net inflow data are constructed, which are:

[0024]

[0025] Where: n is 26; m is 12; S t Represents the data slice of net inflow of northbound funds in the tth interval;

[0026] Get the difference between the slow exponential moving average and the fast exponential moving average to get the DIFF line, which is:

[0027]

[0028] Perform exponential smoothing on the DIFF line to obtain the Dea line, which is:

[0029]

[0030] Where: p is 9;

[0031] Get the difference between the Dea line and the DIFF line to get the constructed NFMACD indicator, which is:

[0032] NFMACD t =NFDiff t -NFDea t .

[0033] The embodiment of the present invention also provides a device for establishing a NFMACD indicator of northbound funds with advanced characteristics, comprising:

[0034] The data module is used to obtain the technical factor indicator data of the net inflow data of the daily northbound funds at the closing time, and use the create function to splice the technical factor indicator data into a multivariate time series data set;

[0035] The model module is used to build a deep learning network model. It takes the daily multivariate time series data set as input and the net inflow data of northbound funds at the closing of the next day as the target, trains the deep learning network model and obtains a prediction model.

[0036] The reconstruction module is used to predict the net inflow data of northbound funds at closing time through the prediction model to obtain the predicted closing value; the predicted closing value is combined with the historical net inflow closing value to reconstruct the daily net inflow data of northbound funds at closing time;

[0037] A construction module is used to simulate the construction method of the MACD indicator based on the reconstructed daily northbound fund net inflow data at closing time to construct the northbound fund NFMACD indicator with advanced characteristics.

[0038] The embodiment of the present invention provides a method and device for establishing a NFMACD indicator of northbound funds with advanced characteristics. Compared with the prior art, the beneficial effects thereof are as follows:

[0039] The present invention firstly uses the daily northbound fund net inflow data to calculate the technical factor indicator data respectively, and splices these data into a multivariate time series data set to construct a deep learning model, takes the multivariate time series data set as input, takes the net inflow data of the next day, i.e., t+1 day, as the target, trains the deep learning model, and then combines the prediction results of the prediction model with the historical data, and adopts the calculation method of the MACD indicator to construct the NFMACD indicator with advanced characteristics; the constructed NFMACD indicator is more advanced than the traditional MACD indicator, so as to solve the lag of the traditional MACD indicator, better assist investment, and greatly improve the accuracy of the judged market trend; at the same time, the constructed NFMACD indicator has achieved a significant improvement in both the rate of return and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the overall process of a method for establishing a NFMACD indicator for northbound funds with advanced characteristics provided by an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of the deep learning model fitting effect of a method for establishing the NFMACD indicator of northbound funds with advanced characteristics provided by an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of a strategy based on the NFMACD indicator for a method for establishing a NFMACD indicator for northbound funds with advanced characteristics provided by an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of the traditional MACD strategy profit trend of a method for establishing the NFMACD indicator of northbound funds with advanced characteristics provided by an embodiment of the present invention;

[0044] Figure 5A schematic diagram of the trend of returns of the leading northbound funds NFMACD strategy for a method for establishing a northbound funds NFMACD indicator with leading characteristics provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0046] See also Figure 1 The embodiment of the present invention provides a method for establishing a NFMACD indicator of northbound funds with advanced characteristics, comprising the following steps:

[0047] Step 1: Use the net inflow data of northbound funds at the closing of each day to calculate the technical factor indicator data, and then splice the technical factor indicator data into a multivariate time series data set.

[0048] Step 2: Build a deep learning model, take the multivariate time series data set as input and the net inflow data on day t+1 as the target, train the deep learning model and obtain the prediction model.

[0049] Step 3: Predict the net inflow data of northbound funds through the forecasting model, combine the predicted value with the historical net inflow value, and reconstruct the daily net inflow data.

[0050] Step 4: Based on the reconstruction of daily net inflow data, construct the NFMACD indicator of northbound funds with advanced characteristics.

[0051] In step one, the technical factor indicator data calculated using the daily northbound net inflow data include: long-short index, moving average difference, simple moving average, index average, exponential smoothing average, momentum index, price oscillation index, deviation rate, momentum index, divergence momentum index, rate of change index, divergence change rate index, SRMIMI modified index and standard deviation.

[0052] In step 2, the deep learning model includes the first fully connected layer (4 neurons), the second LSTM recurrent network (4 units), and a fully connected output layer, while using the SGD optimizer and the loss function RMSE.

[0053] In step three, the daily net inflow data series is reconstructed, which can be expressed as:

[0054] new Northbound Funds(t,z)=[S(tz),···,S(t-1),S(t),P(t+1)]

[0055] Among them: t represents the date of the day; S(t) represents the net inflow of northbound funds on day t; P(t+1) represents the forecast data on day t; z represents the exponential smoothing period.

[0056] In step 4, the construction process of the NFMACD indicator is:

[0057] First, calculate the slow and fast EMA lines of the net inflow of northbound funds respectively. The formula is:

[0058]

[0059] Where: n is 26; m is 12; S t Represents the data slice of net inflow of northbound funds in the tth interval.

[0060] Then the DIFF line is obtained by calculating the difference between the two lines. The formula is:

[0061]

[0062] Then perform exponential smoothing calculation on the DIFF line to obtain the Dea line, which is:

[0063]

[0064] Among them: p is 9.

[0065] Finally, by calculating the difference between the Dea line and the DIFF line, we get the MACD line, which is:

[0066] NFMACD t =NFDiff t -NFDea t

[0067] The specific embodiments are:

[0068] Step 1: Use the daily northbound fund net inflow data of the A-share market to calculate the technical factor indicator data, and splice the factor indicator data into a multivariate time series data set with a time step of 5; Taking into account the implementation time point of the Stock Connect policy and the time window for northbound funds to accumulate to a certain scale, the time period from January 2017 to December 2023 was mainly selected, covering a total of 1,722 trading days of data; As shown in Table 1, a fragment of the net flow data of northbound funds is shown, which is the data for October 2023.

[0069] Table 1 Northbound capital net flow data

[0070] Trading Day Net inflow of northbound funds 20231009 -7460.23 20231010 -5471.27 20231011 -389.96 20231012 6615.75 20231013 -6438.54 20231016 -6476.94 20231017 -3057.5 20231018 -1160.64 20231019 -11704.27 20231020 -1646.22 20231024 -5045.57 20231025 -1295.2 20231026 1226.75 20231027 4668.03 20231030 -2397.61 20231031 -4753.39

[0071] The technical factor data include: short index, moving average difference, simple moving average, exponential average, exponential smoothing difference average, dynamic index, price oscillation index, deviation rate, momentum index, difference difference dynamic index, rate of change index, difference difference change rate index, SRMIMI correction index and standard deviation; the specific introduction of technical factors is shown in Table 2.

[0072] Table 2 Technical factors

[0073]

[0074]

[0075]

[0076] Apply the create_df(stock_list,delta,num) function to integrate the technical factors calculated above into a complete multivariate time series data set. stock_list represents the stock pool code, delta represents the number of days from the benchmark date, and num represents the number of days of data that need to be retrieved.

[0077] The data is then standardized: the StandardScaler() function is used to calculate the mean and standard deviation of the features in the training set, and each feature is independently centered and scaled; then, the calculated mean and standard deviation are stored so that the same proportions can be used for feature scaling on the test set.

[0078] Step 2: Build a deep learning model, take the multivariate time series data set as input and the net inflow data on day t+1 as the target, train the deep learning model and obtain the prediction model.

[0079] The key part of the deep learning model used is the LSTM neural network; LSTM is a variant of a recurrent neural network (RNN) that is specifically designed to solve the modeling and prediction problems of sequence data; compared to traditional RNNs, LSTM can better handle long-term dependencies and control the flow and retention of information through gating mechanisms (such as input gates, forget gates, and output gates), thereby effectively learning and memorizing long-term patterns and dependencies in sequences.

[0080] The main formula and gating mechanism of LSTM neural network are expressed as:

[0081] ① Forget gate: determines how much memory of the previous moment is retained at the current moment; its formula is:

[0082] f t =σ(W f ·[ht-1 ,x t ]+b f )

[0083] Where: f t represents the output of the forget gate; σ represents the activation function (usually sigmoid); W f and b f Represent the weight matrix and bias respectively; represents the hidden state of the previous moment; represents the input of the current moment.

[0084] ② Input gate: determines how much information is currently input and stored in the memory unit; its formula is:

[0085] i t =σ(W i ·[h t-1 , x t ]+b i )

[0086] Where: i t Represents the output of the input gate.

[0087] ③Candidate memory unit: Generate new candidate memory content; the formula is:

[0088]

[0089] in: Represents the output of the candidate memory unit; represents the hyperbolic tangent function.

[0090] ④Memory unit update: Update the state of the memory unit; the formula is:

[0091]

[0092] Where: C t-1 Indicates the state of the memory unit at the previous moment; C t Indicates the state of the memory unit at the current moment.

[0093] ⑤Output gate: determines the hidden state of the current output; its formula is:

[0094] σ t =σ(W o ·[h t-1 ,x t ]+b o )

[0095] Among them: t Represents the output of the output gate.

[0096] ⑥Hidden state update: Calculate the hidden state at the current moment; the formula is:

[0097] h t =o t *tanh(C t )

[0098] Where: h t Indicates the hidden state at the current moment.

[0099] Therefore, LSTM finely controls the flow and retention of information through these gating mechanisms, thereby being able to effectively capture long-term dependencies in time series.

[0100] The construction of the deep learning model includes: the first layer is a fully connected layer with 4 neurons; the second layer is a 4-unit LSTM, and the last is a fully connected output layer, using the SGD optimizer and RMSE loss function.

[0101] After the model is built, the model indicator parameters are determined and optimized, and the final confirmed model indicator parameters are shown in Table 3.

[0102] Table 3 Model index parameters

[0103] Learning Rate Time step Optimizer Loss Function Training cycle batch 0.01 5 SGD RMSE 100 500

[0104] In order to quantitatively evaluate the prediction performance of the model, the root mean square error (RMSE) on the test set and the training set was calculated; the RMSE value on the training set was 0.1175, showing that the model fits the training data well, as shown in Figure 2. Figure 2 As shown in the figure, the RMSE value on the test set is 0.1214, indicating that although there are errors in the model when predicting unknown data, the overall performance is still acceptable.

[0105] Step 3: Predict the net inflow of northbound funds through the forecasting model, combine the predicted net inflow value with the historical net inflow value, and reconstruct the daily net inflow data.

[0106] The reconstructed daily net inflow data is expressed as:

[0107] new Northbound Funds(t,z)=[S(tz),···,S(t-1),S(t),P(t+1)]

[0108] Among them: t represents the date of the day; S(t) represents the net inflow of northbound funds on day t; P(t+1) represents the forecast data on day t; z represents the exponential smoothing period.

[0109] Step 4: Based on the reconstruction of daily net capital inflow data, construct the NFMACD indicator with advanced characteristics.

[0110] The construction process of the NFMACD indicator is:

[0111] First, calculate the slow and fast EMA lines of the net inflow of northbound funds respectively. The formula is:

[0112]

[0113] Where: n is 26; m is 12; S t Represents the data slice of net inflow of northbound funds in the tth interval.

[0114] Then the DIFF line is obtained by calculating the difference between the two lines. The formula is:

[0115]

[0116] Then perform exponential smoothing calculation on the DIFF line to obtain the Dea line, which is:

[0117]

[0118] Among them: p is 9.

[0119] Finally, by calculating the difference between the Dea line and the DIFF line, we get the MACD line, which is:

[0120] NFMACD t =NFDiff t -NFDea t

[0121] In order to verify the improvement of the NFMACD indicator constructed by the present invention compared with the traditional MACD indicator, the operation of the NFMACD indicator and the traditional MACD indicator on the Shanghai Stock Exchange 50 Index are used to test the performance of the two strategies under a market with a large fluctuation range; the strategy design framework is as follows Figure 3 As shown; its main steps are:

[0122] 1. Traditional MACD trading strategy:

[0123] Starting from the constituent stocks of the Shanghai Composite 50, an initial stock pool is constructed; after data preprocessing to ensure that the stock codes and weights are accurate, the strategy enters the core stock selection process.

[0124] First, calculate the MACD value of each stock one by one with a short-term period of 12, a fast period of 26, and a signal line period of 9, that is, MACD(12,26,9). If the MACD(12,26,9) of a stock on that day is greater than 0 and the MACD(12,26,9) of the previous day is less than 0, it will be included in the new stock pool.

[0125] The strategy then focuses on each stock in the new stock pool and calculates its one-way volatility difference and its 60-day moving average; if the 60-day volatility difference average is greater than 0, it is generally believed that the market will continue to rise; conversely, if the 60-day volatility difference average is less than 0, it is believed that the market will continue to fall.

[0126] 2. Strategy based on NFMACD indicator:

[0127] When the fast line of NFMACD crosses the slow line to form a golden cross, and the histogram turns from negative to positive, it can be regarded as a buy signal; on the contrary, when the fast line crosses the slow line downward to form a dead cross, and the histogram turns from positive to negative, it can be regarded as a sell signal; at the same time, in order to effectively control risks, reasonable stop loss points and take profit points are set to ensure that stop loss or take profit can be made in time when the market fluctuates.

[0128] Based on the moving average and upper and lower tracks of NFBBAND, the NFBBAND ​​quantitative investment strategy is constructed: when the current net inflow of northbound funds crosses the upper track of NFBBAND, the quantitative system turns on the switch and executes the traditional MACD timing strategy; when the current net inflow of northbound funds crosses the lower track of NFBBAND, the quantitative system turns off the switch and does not execute the MACD timing strategy.

[0129] like Figure 4 As shown in the figure, the Sharpe ratio of the traditional MACD stock selection strategy is 0.10, which is lower than 1, and the maximum drawdown is 21.94%, which means that the risk of this strategy is relatively high and the return volatility is relatively large. Northbound funds are usually regarded as a market vane, and their flow can reflect the views of foreign capital on the Chinese market. Therefore, the forecast results of northbound funds are used to construct an advanced NFMACD indicator to decide whether to open a position.

[0130] like Figure 5 As shown, the Sharpe ratio of the NFMACD stock selection strategy for leading northbound funds is 1.47, which is higher than 1, and the maximum drawdown is 11.58%. The strategy return is 33.34%, which is much higher than the benchmark return of -8.36%. The benchmark return is -8.36%, which shows that the overall market performed poorly during this period, but the NFMACD strategy was still able to achieve positive returns, further verifying its effectiveness.

[0131] The NFMACD indicator established by the present invention integrates the advantages of the traditional MACD indicator, which is more effective than the method of relying solely on northbound capital flow data to judge market trends; it not only reflects market dynamics in advance, but also improves the rate of return and stability; at the same time, it effectively solves the lag problem of the traditional MACD indicator, thereby more effectively assisting investment decisions.

[0132] The present invention firstly uses the daily northbound fund net inflow data to calculate the technical factor indicator data respectively, and splices these data into a multivariate time series data set to construct a deep learning model. The multivariate time series data set is used as input and the t+1 day net inflow data is used as the target to train the deep learning model. Then the prediction results of the prediction model will be combined with the historical data, and the calculation method of the MACD indicator will be used to construct the NFMACD indicator with advanced characteristics. This indicator integrates the advantages of the traditional MACD and is more effective in predicting market trends than simply relying on the northbound fund flow data. It is more advanced than the traditional MACD indicator.

[0133] The embodiment of the present invention also provides a device for establishing a NFMACD indicator of northbound funds with advanced characteristics, comprising:

[0134] The data module is used to use the net inflow data of northbound funds at the closing time of each day to obtain the technical factor indicator data of the net inflow data, and use the create function to splice the technical factor indicator data into a multivariate time series data set.

[0135] The model module is used to build a deep learning network model. It takes the daily multivariate time series data set as input and the net inflow data of northbound funds at the closing of the next day as the target. It trains the deep learning network model to obtain a prediction model.

[0136] The reconstruction module is used to predict the net inflow data of northbound funds at closing time through the prediction model to obtain the predicted closing value; combine the predicted closing value with the historical net inflow closing value to reconstruct the daily net inflow data of northbound funds at closing time.

[0137] A construction module is used to simulate the construction method of the MACD indicator based on the reconstructed daily northbound fund net inflow data at closing time to construct the northbound fund NFMACD indicator with advanced characteristics.

[0138] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for establishing the NFMACD indicator of northbound funds with advanced characteristics, characterized in that: The following steps are involved: Use the daily northbound fund closing net inflow data to obtain the technical factor indicator data of the net inflow data, and use the create function to splice the technical factor indicator data into a multivariate time series data set; Construct a deep learning network model, take the daily multivariate time series data set as input, and take the net inflow data of northbound funds at the closing of the next day as the target, train the deep learning network model and obtain a prediction model; Predict the net inflow data of northbound funds at closing time through the prediction model to obtain the predicted closing value; combine the predicted closing value with the historical net inflow closing value to reconstruct the daily net inflow data of northbound funds at closing time; Based on the reconstructed daily northbound fund net inflow data at closing, the construction method of the MACD indicator is simulated to construct the northbound fund NFMACD indicator with advanced characteristics.

2. A method for establishing the NFMACD indicator of northbound funds with advanced characteristics according to claim 1, characterized in that: The technical factor indicator data include: Long-short index, moving average difference, simple moving average, exponential average, exponential smoothing difference, momentum index, price oscillator index, deviation rate, momentum index, difference difference momentum index, rate of change index, difference difference change rate index, SRMIMI modified index and standard deviation.

3. A method for establishing the NFMACD indicator of northbound funds with advanced characteristics according to claim 1, characterized in that: After obtaining the multivariate time series data set, the principal component analysis method is used to perform dimensionality reduction processing on the multivariate time series data set.

4. A method for establishing the NFMACD indicator of northbound funds with advanced characteristics according to claim 1, characterized in that: The deep learning network model includes a fully connected input layer, an LSTM recurrent network and a fully connected output layer; The number of neurons in the fully connected input layer is 4; The number of units in the LSTM recurrent network is 4; The weight optimizer of the deep learning network model is an SGD optimizer; The loss function of the deep learning network model is RMSE.

5. The method for establishing the NFMACD indicator of northbound funds with advanced characteristics according to claim 1 is characterized in that: The reconstructed daily northbound funds closing net inflow data is: new Northbound Funds(t,z)=[S(tz),···,S(t-1),S(t),P(t+1)] Among them: t represents the date of the day; S(t) represents the net inflow of northbound funds on day t; P(t+1) represents the forecast data on day t; z represents the exponential smoothing period.

6. A method for establishing the NFMACD indicator of northbound funds with advanced characteristics according to claim 1, characterized in that: The construction of the NFMACD indicator includes: Based on the reconstructed daily northbound fund closing net inflow data, the slow exponential moving average EMA and fast exponential moving average EMA of the northbound fund net inflow data are constructed, which are: Among them: n is 26; m is 12; S t Represents the data slice of net inflow of northbound funds in the tth interval; Get the difference between the slow exponential moving average and the fast exponential moving average to get the DIFF line, which is: Perform exponential smoothing on the DIFF line to obtain the Dea line, which is: Where: p is 9; Get the difference between the Dea line and the DIFF line to get the constructed NFMACD indicator, which is: NFMACD t =NFDiff t -NFDea t 。 7. A device for establishing the NFMACD indicator of northbound funds with advanced characteristics, characterized in that: include: The data module is used to obtain the technical factor indicator data of the net inflow data of the daily northbound funds at the closing time, and use the create function to splice the technical factor indicator data into a multivariate time series data set; The model module is used to build a deep learning network model. It takes the daily multivariate time series data set as input and the net inflow data of northbound funds at the closing of the next day as the target, trains the deep learning network model and obtains a prediction model. The reconstruction module is used to predict the net inflow data of northbound funds at closing time through the prediction model to obtain the predicted closing value; the predicted closing value is combined with the historical net inflow closing value to reconstruct the daily net inflow data of northbound funds at closing time; A construction module is used to simulate the construction method of the MACD indicator based on the reconstructed daily northbound fund net inflow data at closing time to construct the northbound fund NFMACD indicator with advanced characteristics.

Images