Financial securities product predicting system based on convolution neural network

Abstract

This invention provides a financial securities product prediction system utilizing convolutional neural networks (CNNs). The system includes a cloud database and a processor. The processor is configured to retrieve transaction data of financial securities products from the cloud database; analyze the transaction data to generate multiple parameter values ​​for each item in the transaction data; divide the financial securities products into multiple groups based on the parameter values ​​of the transaction data; build a prediction model using the CNN based on the parameter values ​​corresponding to each group; input the parameter values ​​of one financial securities product into the prediction model to predict one of the groups corresponding to the financial securities products; and generate an investment strategy based on one of the groups corresponding to the financial securities products. This assists investors with different characteristics in making suitable investment decisions.

Description

Financial Securities Product Prediction System Using Convolutional Neural Networks financial securities product prediction system based on convolution neural network Technical Field

[0001] This invention relates to a financial securities product prediction system, and more particularly to a financial securities product prediction system that utilizes convolutional neural networks. Prior Technology

[0002] With the rapid development of artificial intelligence technology, its application in the financial field has gradually become widespread. However, due to the complexity and dynamism of the financial market, it is difficult to accurately predict changes in the financial market.

[0003] In view of this, developing a financial securities product prediction system using convolutional neural networks that can provide the best investment strategies based on different types of consumers has become a worthwhile research and development goal for relevant industry players. Summary of the Invention

[0004] Therefore, the purpose of this invention is to provide a financial securities product prediction system that utilizes convolutional neural networks (CNNs) to generate suitable investment strategies by analyzing transaction data of financial securities products. In this way, this novel financial securities product prediction system using CNNs automatically learns and captures the complex characteristics related to financial securities products, assisting consumers in understanding and predicting the financial market.

[0005] According to one embodiment of the present invention, a financial securities product prediction system using convolutional neural networks is provided, comprising a cloud database and a processor. The cloud database contains multiple transaction data for multiple financial securities products. The processor is signal-connected to the cloud database and configured to perform operations including the following steps: obtaining the transaction data from the cloud database; analyzing the transaction data to generate multiple parameter values ​​corresponding to each of the transaction data; dividing the financial securities products into multiple groups based on the parameter values ​​of the transaction data; establishing a prediction model through a convolutional neural network based on the parameter values ​​corresponding to the groups; inputting the parameter values ​​of one of the financial securities products into the prediction model to predict one of the groups corresponding to that financial securities product; and generating an investment strategy based on one of the groups corresponding to that financial securities product.

[0006] Other embodiments of the aforementioned implementation method are as follows: one of the aforementioned transaction data includes a financial securities product code, a date, a trading volume, a price increase, a total amount, and a closing price.

[0007] Other embodiments of the aforementioned implementation are as follows: The aforementioned parameter values ​​include a risk factor, a growth value factor, a size factor, a liquidity factor, and a momentum factor.

[0008] Other embodiments of the foregoing implementation are as follows: The foregoing processor further includes operations configured to perform the following steps: adjusting at least one of the following: the number of multiple convolutional layers, the number of multiple pooling layers, the number of multiple hidden neurons, the number of multiple filters, the size of each of these filters, and the size of each of these pooling layers.

[0009] Other embodiments of the aforementioned implementation are as follows: The aforementioned financial securities product prediction system using convolutional neural networks further includes a display. The display is signal-connected to the processor and is used to display one of the groups and investment strategies corresponding to these financial securities products. Simple Explanation of the Diagram

[0010] Figure 1 is a block diagram illustrating a financial securities product prediction system using convolutional neural networks according to a first embodiment of the present invention; Figure 2 is a flowchart illustrating the second embodiment of the present invention, which describes a financial securities product prediction method using convolutional neural networks; Figure 3 is a block diagram illustrating a financial securities product prediction system using convolutional neural networks according to a third embodiment of the present invention; and Figure 4 is a flowchart illustrating the fourth embodiment of the present invention, which describes a financial securities product prediction method using convolutional neural networks. Implementation

[0011] Several embodiments of the present invention will be described below with reference to the drawings. For clarity, many practical details will be set forth in the following description. However, it should be understood that these practical details are not intended to limit the present invention. That is, in some embodiments of the present invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and elements will be shown in a simple schematic manner in the drawings; and repeated elements may be denoted by the same number.

[0012] Furthermore, in this document, when a component (or unit or module, etc.) is "connected" to another component, it can mean that the component is directly connected to the other component, or that the component is indirectly connected to the other component, meaning that there is another component between the component and the other component. Only when it is explicitly stated that a component is "directly connected" to another component does it indicate that there is no other component between the component and the other component. The terms "first," "second," and "third" are only used to describe different components and do not limit the components themselves; therefore, "first component" can also be referred to as "second component." Moreover, the combinations of components / units / circuits in this document are not combinations generally known, conventional, or customary in this field. Whether the components / units / circuits themselves are customary cannot be used to determine whether their combination relationships are easily accomplished by someone with ordinary knowledge in the technical field.

[0013] Please refer to Figures 1 and 2. Figure 1 is a block diagram illustrating the financial securities product prediction system 100 using convolutional neural networks according to the first embodiment of this invention; Figure 2 is a flowchart illustrating the financial securities product prediction method S10 using convolutional neural networks according to the second embodiment of this invention. The financial securities product prediction system 100 using convolutional neural networks includes a cloud database 110 and a processor 120. The cloud database 110 contains multiple transaction data D1 of multiple financial securities products. The processor 120 is signal-connected to the cloud database 110 and configured to implement the financial securities product prediction method S10 using convolutional neural networks. The financial securities product prediction method S10 using convolutional neural networks includes steps S01, S02, S03, S04, S05, and S06.

[0014] Step S01 involves the driver processor 120 retrieving transaction data D1 of these financial securities products from the cloud database 110. Step S02 involves the driver processor 120 analyzing the transaction data D1 to generate a complex parameter value P1 corresponding to each item in the transaction data D1. Step S03 involves the driver processor 120 dividing these financial securities products into multiple groups based on the parameter values ​​P1 of the transaction data D1. Step S04 involves the driver processor 120 establishing a prediction model M1 using a convolutional neural network based on the parameter values ​​P1 corresponding to each group. Step S05 involves the driver processor 120 inputting the parameter values ​​P1 of each of these financial securities products into the prediction model M1 to predict one of the groups corresponding to each of these financial securities products. Step S06 involves the driver processor 120 generating an investment strategy D2 based on one of the groups corresponding to each of these financial securities products.

[0015] In detail, the cloud database 110 may be memory or other storage device, and the processor 120 may be a central processing unit (CPU), a virtual private server (VPS), or other electronic computing device; this invention is not limited thereto.

[0016] In step S01, the transaction data D1 obtained by the processor 120 from the cloud database 110 represents the transaction status of all financial securities products within a specific time period. This data may include the financial securities product code, date, trading volume, price increase, total amount, closing price, and price increase within a given period for each financial securities product; however, this invention is not limited to these parameters. The transaction data D1 may be as shown in Table 1. Table 1 date March 21, 2024 Financial securities products code Trading volume Increase Total closing price Range of increase Taiwan Cement 1101 32 -0.47 17833 31.5 1.58 Asia Cement 1102 42.5 0.12 2749 41.35 2.78 Jia Ni 1103 17.85 0.28 280 17.2 3.78 Ring mud 1104 33.85 0.45 1120 32.25 4.96

[0017] In detail, the parameter value P1 may include a risk factor (BETA value), a growth value factor (P / B), a size factor (market capitalization), a liquidity factor (turnover rate), and a momentum factor (daily closing price - transaction price). In step S02, the processor 120 calculates the five factors (i.e., risk factor, growth value factor, size factor, liquidity factor, and momentum factor) for each financial security product based on all transaction data D1 within a specific time period, as shown in Table 2. For example, when analyzing the momentum factor of a financial security product, the processor 120 can first find the value with the highest price change over 12 months and the value with the lowest price change over 12 months, and subtract the two to obtain the momentum factor of this financial security product. Alternatively, it can select the 100 transaction data D1 with the highest values ​​from all transaction data D1 within a specific time period for analysis to obtain the result, but this invention is not limited to these methods. Table 2 Financial securities products code Risk factors Growth Value Factor Size factor Liquidity factor Kinetic energy factor Taiwan Cement 1101 1 1 2379 9 17 Asia Cement 1102 0 1 1467 5 9 Jia Ni 1103 1 1 136 2 19 Ring mud 1104 0 1 217 8 19

[0018] In step S03, the processor 120 maps the parameter values ​​P1 of each financial security product to a two-dimensional space using a self-organizing map network (Kohonen), visually mapping and analyzing the data. Based on the difference between the landing point of the parameter value P1 in the two-dimensional space and other data, these financial security products are divided into multiple groups. In this embodiment, the number of groups is three, which can be a high-risk group, a medium-risk group, and a low-risk group. Step S03 can further calculate the importance of the parameter value P1 to the grouping results. The risk factor has the greatest impact on the grouping results, while the kinetic factor has the second greatest impact, but this invention is not limited to this.

[0019] In step S04, the financial securities products in the three groups and their corresponding parameter values ​​P1 are divided into training set data and test set data and input into the convolutional neural network for training and prediction, thereby generating the prediction model M1.

[0020] In step S05, after the prediction model M1 is established, the financial securities products that have not been grouped in step S03 and their corresponding parameter values ​​P1 can be input into the prediction model M1 to confirm the group to which the aforementioned financial securities products belong.

[0021] In step S06, the processor 120 can generate an investment strategy D2 based on whether each financial security product corresponds to a high-risk, medium-risk, or low-risk group, and in conjunction with its parameter value P1. Furthermore, the level of the parameter value P1 corresponding to each financial security product is also a basis for the processor 120 to generate the investment strategy D2. For example, financial security products with higher risk factor and growth value factor values ​​typically have higher volatility and can therefore be recommended to consumers with high-risk characteristics.

[0022] Therefore, the novel financial securities product prediction method S10 using convolutional neural networks can identify financial securities products suitable for risk groups of consumers with different characteristics through clustering, and then capture the complex features related to financial securities products through automatic learning, thereby assisting consumers in understanding and predicting the financial market.

[0023] Please refer to Figures 1 through 3, where Figure 3 is a block diagram illustrating a financial securities product prediction system 200 using convolutional neural networks according to a third embodiment of the present invention. The financial securities product prediction system 200 using convolutional neural networks includes a cloud database 210, a processor 220, and a display 230. The cloud database 210 contains multiple transaction data D1 of multiple financial securities products. The processor 220 is signal-connected to the cloud database 210. In the third embodiment, the cloud database 210 and processor 220 of the financial securities product prediction system 200 using convolutional neural networks operate in the same way as the cloud database 110 and processor 120 of the financial securities product prediction system 100 using convolutional neural networks in the first embodiment, and will not be described again. Specifically, the financial securities product prediction system 200 using convolutional neural networks may further include a display 230. The display 230 is signal-connected to the processor 220 and is used to display a group of financial securities products and an investment strategy D2 corresponding to them. In this way, the novel financial securities product prediction system 200 using convolutional neural networks displays investment strategies D2 (such as recommended financial securities products and the nature of each financial securities product) that match the consumer's consumption characteristics on the display 230, allowing the consumer to review and refer to suitable investment strategies D2.

[0024] Please refer to Figures 1 through 4, where Figure 4 is a flowchart illustrating the financial securities product prediction method S10a using convolutional neural networks according to the fourth embodiment of this invention. The financial securities product prediction system 200 using convolutional neural networks can be configured to implement the financial securities product prediction method S10a using convolutional neural networks. The financial securities product prediction method S10a using convolutional neural networks includes steps S11, S12, S13, S14, S15, S16, S17, and S18. In the fourth embodiment, steps S11, S12, S13, S14, S16, and S17 of the financial securities product prediction method S10a using convolutional neural networks are identical to steps S01, S02, S03, S04, S05, and S06 of the financial securities product prediction method S10 using convolutional neural networks, and will not be described again. Specifically, the financial securities product prediction method S10a using convolutional neural networks may further include steps S15 and S18. Step S15 includes driving the processor 220 to adjust at least one of the following: the number of multiple convolutional layers, the number of multiple pooling layers, the number of multiple hidden neurons, the number of multiple filters, the size of each filter, and the size of each pooling layer.

[0025] In detail, the convolutional neural network includes multiple convolutional layers, multiple pooling layers, and fully connected layers. In the fourth embodiment, the convolutional neural network includes two convolutional layers, two pooling layers, and the number of filters in the two convolutional layers are 16 and 36, respectively. The sizes of the two filters are (3,3) and (5,5), respectively. The sizes of the two pooling layers are (2,2) and (1,1), respectively. The number of hidden neurons is 128 and 3, respectively. The accuracy of the prediction model M1 generated by the aforementioned convolutional neural network is 0.8202.

[0026] In step S15, the accuracy of prediction model M1 can be improved by changing the number of convolutional layers, the number and size of pooling layers, the number of hidden neurons, or the number and size of filters. In the fourth embodiment, when the convolutional neural network adds one convolutional layer and one pooling layer, the accuracy of prediction model M1 can be improved to 0.831; when the convolutional neural network adds one convolutional layer, the accuracy of prediction model M1 can be improved to 0.8338; when the number of filters in the second convolutional layer of the convolutional neural network is adjusted to 20, the accuracy of prediction model M1 can be improved to 0.8283; when the size of the pooling layer of the convolutional neural network is adjusted to (3,3), the accuracy of prediction model M1 can be improved to 0.8256; when the number of hidden neurons in the first convolutional layer of the convolutional neural network is adjusted to 256, the accuracy of prediction model M1 can be improved to 0.8365, but the present invention is not limited thereto.

[0027] Therefore, the novel financial securities product prediction method S10a using convolutional neural networks can improve the accuracy of the prediction model M1 by adjusting the structure of the convolutional neural network in step S15.

[0028] Step S18 includes driving the display 230 to display one of the financial securities products and the corresponding group and investment strategy D2.

[0029] As can be seen from the above embodiments, the novel financial securities product prediction method and system using convolutional neural networks has the following advantages: First, it can identify financial securities products suitable for risk groups of consumers with different characteristics through clustering, and then automatically learn and capture the complex characteristics related to financial securities products to assist consumers in understanding and predicting the financial market; Second, it displays investment strategies that match the consumer's consumption characteristics (such as recommended financial securities products and the nature of each financial securities product) on the display screen for consumers to review and refer to suitable investment strategies; Third, it improves the accuracy of the prediction model by adjusting the structure of the convolutional neural network through the steps.

[0030] Although the present invention has been disclosed above with reference to embodiments, it is not intended to limit the present invention. Anyone skilled in the art can make various modifications and alterations without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

[0031] 100,200: A Financial Securities Product Prediction System Utilizing Convolutional Neural Networks 110,210: Cloud-based database 120,220: Processor 230: Monitor D1: Transaction Data D2: Investment Strategy M1: Predictive Model P1: Parameter value S01, S02, S03, S04, S05, S06, S11, S12, S13, S14, S15, S16, S17, S18: Steps S10, S10a: Financial Securities Product Prediction Methods Using Convolutional Neural Networks

Claims

1. A financial securities product prediction system using convolutional neural networks, comprising: a cloud database containing multiple transaction data of multiple financial securities products; and a processor signal-connected to the cloud database and configured to perform operations including the following steps: obtaining the transaction data from the cloud database; analyzing the transaction data to generate multiple parameter values ​​corresponding to each of the transaction data; classifying the financial securities products into multiple groups based on the parameter values ​​of the transaction data; establishing a prediction model through a convolutional neural network based on the parameter values ​​corresponding to the groups; inputting the parameter values ​​of one of the financial securities products into the prediction model to predict one of the groups corresponding to that financial securities product; and generating an investment strategy based on the group corresponding to that financial securities product.

2. The financial security product prediction system using convolutional neural networks as described in claim 1, wherein one of the transaction data includes a financial security product code, a date, a trading volume, a price change, a total amount, and a closing price.

3. The financial securities product prediction system using convolutional neural networks as described in claim 1, wherein the parameter values ​​include a risk factor, a growth value factor, a size factor, a liquidity factor, and a momentum factor.

4. The financial securities product prediction system using convolutional neural networks as described in claim 1, wherein the processor further comprises operations configured to perform the following steps: adjusting at least one of the number of multiple convolutional layers, the number of multiple pooling layers, the number of multiple hidden neurons, the number of multiple filters, the size of each of the filters, and the size of each of the pooling layers of the convolutional neural network.

5. The financial securities product prediction system using convolutional neural networks as described in claim 1 further includes: a display, signal-connected to the processor, for displaying the group corresponding to the financial securities products and the investment strategy.

Images