Intelligent period and current fusion automatic purchasing system and method

Abstract

The invention relates to an intelligent period and current fusion automatic purchasing system and method, and the method comprises the steps: collecting the real-time market data of a futures market and the quotation data of a spot market according to the collection frequency, and carrying out the preprocessing of the collected data; the period-present analysis module calculates key indexes, judges the market state, identifies whether an arbitrage opportunity or abnormal price fluctuation occurs, and marks the arbitrage opportunity if a period-present price difference rate stock investment threshold alpha; the intelligent decision engine generates a purchase strategy according to the market state and the enterprise demand, and performs risk inspection on the generated purchase strategy through the risk control module; and executing a purchase instruction after all the risk checks pass, monitoring the order state in real time, and finally storing the complete record of the purchase. Real-time synchronization and centralized display of futures and spot market data are realized, purchasers can acquire comprehensive market information on a single interface, the information acquisition efficiency is improved by more than 80%, and decision errors caused by information lag are effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of data processing, and in particular to an intelligent futures-spot fusion automated procurement system and method. Background Technology

[0002] With the rapid development of the commodity market, enterprises face an increasingly complex market environment in the process of raw material procurement. The futures market and the spot market, as two interconnected markets with different operating mechanisms, provide enterprises with diversified procurement channels. The futures market has the functions of price discovery and risk hedging, while the spot market can meet the actual material needs of enterprises. Traditional procurement models typically separate futures trading and spot procurement, which are operated independently by different business departments, lacking an effective coordination mechanism. In recent years, with the development of financial technology, some enterprises have begun to try to build procurement management systems to assist in procurement decisions. However, existing procurement systems mainly focus on traditional areas such as order management and supplier management, and have weak capabilities in integrating the futures market and the spot market, making it difficult to achieve intelligent procurement decisions and automated transaction execution.

[0003] Therefore, existing technologies have the following drawbacks: 1. Futures and spot market data are scattered, lacking a unified data collection and integration platform. Procurement personnel need to switch between multiple systems to view information, resulting in low information acquisition efficiency and a tendency to miss key market signals. 2. Procurement decisions rely on human experience and judgment, lacking scientific quantitative analysis tools. Faced with complex market fluctuations, it is difficult to make optimal decisions quickly, leading to significant decision delays and subjective biases. 3. There is a lack of linkage between futures hedging and spot procurement. Procurement strategies cannot be automatically adjusted based on market signals such as futures-spot price differences and basis changes, resulting in poor hedging effectiveness or high procurement costs. 4. The procurement execution process requires manual order placement. During rapid price fluctuations, optimal trading opportunities can be missed, and manual operation carries the risk of errors, potentially causing economic losses. 5. There is a lack of effective risk control mechanisms. Price risk, financial risk, and operational risk cannot be monitored in real time during the procurement process, making it difficult to stop losses in a timely manner in case of anomalies. 6. Procurement data is stored in a scattered manner, making it difficult to effectively utilize historical transaction data and continuously optimize procurement strategies through data mining and machine learning technologies, resulting in a waste of data assets. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent futures and spot fusion automatic procurement system and method, which solves the deficiencies of the prior art.

[0005] The objective of this invention is achieved through the following technical solution: an intelligent futures-spot fusion automatic procurement system, the system comprising a data acquisition module, a data clarification module, a futures-spot analysis module, an intelligent decision engine, a risk control module, and an automatic trading module; The data acquisition module is configured to collect market data in real time from futures exchanges, spot trading platforms, and industry information websites. The data clarification module is configured to perform outlier detection, missing value imputation, and data standardization on the collected raw data. The futures-spot analysis module is configured to use time series analysis algorithms and statistical models to perform real-time analysis of futures price trends, spot price trends, futures-spot price spreads, and basis changes, and generate market trend forecast reports. The intelligent decision engine is configured to automatically generate procurement recommendations based on a multi-factor quantification model and machine learning algorithms, taking into account price factors, inventory factors, capital factors, and risk factors. The risk control module is configured to set multiple risk thresholds and automatically suspend transactions or send alarm information when a risk warning is triggered. The automated trading module is configured to automatically send trading instructions to futures brokers or spot suppliers through a trading interface based on the instructions of the intelligent decision engine, and to track the order execution status in real time.

[0006] The system also includes a data storage module, which is configured to use a distributed database to store historical transaction data, market data, and decision records.

[0007] The intelligent decision-making engine automatically generates procurement recommendations, including: A1. Calculate the futures purchase cost C_futures and spot purchase cost C_spot; A2. Calculate the spot-futures price advantage index K = (C_spot - C_futures) / C_spot × 100%; A3. If K > β, prioritize futures purchases; if K < -β, prioritize spot purchases; if -β ≤ K ≤ β, adopt a mixed purchase strategy, where β is the threshold. A4. Calculate the purchase quantity Q = max(S - I + D, 0) based on the enterprise inventory level I and safety stock S, where D is the predicted demand for the next 30 days. A5. Calculate the hedging ratio H = min(λ × σ / 0.15, 1.0) based on the price volatility σ and the risk preference coefficient λ, where λ is the risk preference coefficient.

[0008] The risk control module performs risk verification on the generated procurement strategy, including: B1. Does the amount of a single transaction exceed the budget limit L1? B2. Does the total transaction amount for the day exceed the daily limit L2? B 3. Does the holding ratio after the purchase exceed the maximum holding ratio R_max? B4. Does the current market volatility exceed the abnormal volatility threshold σ_max? B5. Do the available funds meet the margin requirements?

[0009] A method based on an intelligent futures-spot fusion automated procurement system, the method comprising: S1. Collect real-time market data and spot market price data according to the collection frequency, and preprocess the collected data; S2. The futures-spot analysis module calculates key indicators and judges the market status, identifies whether there are arbitrage opportunities or abnormal price fluctuations. If the futures-spot price difference rate exceeds the stock trading threshold α, it is marked as an arbitrage opportunity. S3, the intelligent decision engine generates procurement strategies based on market conditions and enterprise needs, and performs risk verification on the generated procurement strategies through the risk control module; S4. After all risk checks are passed, execute the procurement order, monitor the order status in real time, and finally store the complete record of this procurement.

[0010] The intelligent decision-making engine generates procurement strategies based on market conditions and enterprise needs, including: A1. Calculate the futures purchase cost C_futures and spot purchase cost C_spot; A2. Calculate the spot-futures price advantage index K = (C_spot - C_futures) / C_spot × 100%; A3. If K > β, prioritize futures purchases; if K < -β, prioritize spot purchases; if -β ≤ K ≤ β, adopt a mixed purchase strategy, where β is the threshold. A4. Calculate the purchase quantity Q = max(S - I + D, 0) based on the enterprise inventory level I and safety stock S, where D is the predicted demand for the next 30 days. A5. Calculate the hedging ratio H = min(λ × σ / 0.15, 1.0) based on the price volatility σ and the risk preference coefficient λ, where λ is the risk preference coefficient.

[0011] The risk verification of the generated procurement strategy through the risk control module includes: B1. Does the amount of a single transaction exceed the budget limit L1? B2. Does the total transaction amount for the day exceed the daily limit L2? B 3. Does the holding ratio after the purchase exceed the maximum holding ratio R_max? B4. Does the current market volatility exceed the abnormal volatility threshold σ_max? B5. Do the available funds meet the margin requirements?

[0012] This invention offers the following advantages: An intelligent futures-spot integrated automated procurement system and method establishes a unified data collection and integration platform, enabling real-time synchronization and centralized display of futures and spot market data. Procurement personnel can access comprehensive market information on a single interface, improving information acquisition efficiency by over 80% and effectively avoiding decision-making errors caused by information lag. It can automatically identify the optimal procurement timing and channels, reducing average procurement costs by 3%-5% compared to manual decision-making, and accelerating decision-making speed from minutes to seconds. It achieves an intelligent linkage mechanism between the futures and spot markets, automatically adjusting procurement strategies based on changes in futures-spot price spreads and basis. When futures prices have a significant advantage, it increases the proportion of futures purchases and simultaneously establishes hedging positions; when spot prices are low, it directly purchases spot goods, improving hedging effectiveness by over 40%. It automates the entire procurement transaction process, from market monitoring, strategy generation, risk control to order execution. While requiring manual intervention, the system reduces transaction execution speed from minutes to milliseconds, enabling timely capture of optimal trading opportunities during periods of rapid price fluctuation and mitigating human error. A multi-layered risk control system is designed, covering multiple dimensions such as transaction amount, position ratio, price volatility, and fund usage. Each transaction undergoes risk verification before execution, and automatic stop-loss is triggered immediately upon reaching a risk threshold, effectively preventing operational and market risks. No major risk events have occurred since the system's inception. A complete data asset management system has been established, with all market, transaction, and decision-making data stored in a structured manner. Procurement strategies are continuously optimized through machine learning algorithms; the longer the system is used, the higher the decision accuracy, creating a positive feedback loop. Flexible operating modes are available, allowing for fully automated operation to improve efficiency, or a semi-automatic mode that retains manual review, meeting the risk control requirements and management habits of different enterprises and demonstrating strong applicability. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a schematic flowchart of the method of the present invention; Figure 3 This is a flowchart of the intelligent decision engine of the present invention; Figure 4 This is a flowchart illustrating the inspection process of the risk control module of this invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application provided below with reference to the accompanying drawings is not intended to limit the scope of protection of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. The present invention will be further described below with reference to the accompanying drawings.

[0015] like Figure 1 As shown, one embodiment of the present invention relates to an intelligent futures-spot fusion automated procurement system, applied to the iron ore procurement business of a large steel enterprise. The system is deployed in the enterprise's private cloud environment, adopting a distributed microservice architecture, and specifically includes the following: The data acquisition module connects to the market data systems of the Shanghai Futures Exchange and the Dalian Commodity Exchange via dedicated lines to obtain real-time data on iron ore futures contracts, including price, trading volume, and open interest, at a frequency of twice per second. Simultaneously, the module connects to the pricing platforms of 10 major spot traders via HTTP interfaces, collecting spot price information every 5 minutes. Furthermore, the module subscribes to the APIs of industry news websites to obtain fundamental data such as inventory levels, port throughput, and mine production.

[0016] The data cleaning module uses the Isolation Forest algorithm for outlier detection, identifying and removing abnormal jumps in price data. For missing spot price data, forward imputation is used to complete the data. All price data is uniformly converted to include tax and adjusted according to standard quality indicators.

[0017] The futures and spot analysis module maintains a real-time calculation engine, updating analysis indicators every minute. Key calculations include: the price difference between the main futures contract and the average spot price, the deviation of the current price difference from the historical average, the price difference structure between contracts of different delivery months, and the number of days of available spot inventory at ports. This module uses the ARIMA model to predict futures price trends for the next 7 days, achieving a prediction accuracy (within 2% error) of 72%.

[0018] The intelligent decision-making engine is the core component of the system, incorporating a deep reinforcement learning model trained on six months of historical data. This engine runs a complete decision-making process every hour, generating procurement recommendations. The decision-making process considers the following factors: current inventory level is 80,000 tons, safety stock is set at 100,000 tons, and the estimated consumption over the next 30 days is 150,000 tons, therefore requiring the procurement of 170,000 tons of iron ore; the current futures price is 680 yuan / ton, plus a delivery cost of 120 yuan / ton, bringing the total cost of futures procurement to 800 yuan / ton; the average spot price is 825 yuan / ton; the futures-spot price spread is (825-800) / 825=3.03%, exceeding the 3% threshold, triggering a futures procurement priority strategy. Simultaneously, considering the current relatively low price, the model recommends a procurement ratio of 70% futures + 30% spot, with a 50% hedging ratio for the futures position.

[0019] The risk control module is configured with the following risk parameters: a maximum single transaction amount of 50 million yuan, a daily transaction amount of 100 million yuan, a maximum position size not exceeding three months' worth of demand, a daily volatility threshold exceeding 5%, and available funds must cover 150% of the margin requirement. The futures portion of this procurement decision amounted to 6.8 million × 119,000 = 80.92 million yuan, exceeding the single transaction limit. Therefore, the system automatically split this procurement into two transactions. The spot procurement amounted to 8.25 million × 51,000 = 42.075 million yuan, which did not trigger any risk restrictions.

[0020] After passing the risk assessment, the automated trading module sent buy orders to the futures company's CTP trading system, purchasing 60,000 tons and 59,000 tons of the main futures contract in two batches, using market orders to ensure execution. Simultaneously, purchase orders were sent to three spot suppliers, each ordering 17,000 tons, requiring delivery in batches within 10 days. All orders were executed within 5 minutes, with an average futures transaction price of 682 yuan / ton and an average spot transaction price of 823 yuan / ton.

[0021] The data storage module uses PostgreSQL to store structured data, MongoDB to store unstructured market information data, and InfluxDB to store time-series market data.

[0022] The user management interface is developed using the Vue.js framework, providing a PC web interface and a mobile app. Purchasing managers can use the interface to view the system's operational status in real time, review pending procurement suggestions, query historical transaction records, and analyze procurement performance. The interface shows that a total of 520,000 tons were procured this month, with an average procurement cost of 813 yuan / ton, 2.8% lower than the market average, resulting in cost savings of approximately 15 million yuan.

[0023] like Figure 2As shown, another embodiment of the present invention relates to an intelligent futures-spot fusion automatic procurement method, which specifically includes the following: Step 1: System initialization, loading user configuration parameters, including procurement varieties, budget amount, risk preference, hedging strategy, etc. Step 2: The data acquisition module starts up, collecting real-time market data for the futures market every 1 second and spot market price data every 5 minutes. Step 3: The data cleaning module preprocesses the collected data, removes abnormal data, and fills in missing data using linear interpolation. Step 4: The futures-spot analysis module calculates key indicators, including the futures-spot price spread, basis, price volatility, and volume change rate. Step 5: The futures-spot analysis module determines the market status and identifies whether there are arbitrage opportunities or abnormal price fluctuations. If the futures-spot price difference exceeds the threshold α (usually set to 3%), it is marked as an arbitrage opportunity. Step 6, as follows Figure 3 As shown, the intelligent decision engine generates procurement strategies based on market conditions and enterprise needs. Specifically, the intelligent decision engine uses deep reinforcement learning (DRL) as its core driving mechanism, modeling procurement decisions as a Markov decision process (MDP): It uses multi-dimensional market states such as futures prices, spot prices, trading volume, inventory levels, historical price series, and fundamental indicators as inputs to the state space, and four action types—"buy futures," "buy spot," "mixed procurement," and "no procurement for now"—as the action space. It continuously optimizes procurement strategy parameters by maximizing long-term cumulative rewards (i.e., minimizing overall procurement costs while ensuring inventory safety). The decision rules described in steps 6.1 to 6.5 below are the specific manifestation of the optimal strategy output by the deep reinforcement learning model after training convergence in the execution phase. The model's training mechanism (state space, action space, reward function, network structure, training process) is explained in detail in steps I to VI below. These two parts together constitute the complete intelligent decision engine. 6.1 Calculate the futures purchase cost C_futures and the spot purchase cost C_spot; 6.2 Calculate the spot-futures price spread advantage index K = (C_spot - C_futures) / C_spot × 100%; 6.3 If K > β (usually set to 2%), futures purchases are preferred; if K < -β, spot purchases are preferred; if -β ≤ K ≤ β, a mixed purchase strategy is adopted. 6.4 Calculate the purchase quantity Q = max(S - I + D, 0) based on the enterprise's inventory level I and safety stock S, where D is the predicted demand for the next 30 days; 0 is the non-negative lower limit of the purchase quantity - when the existing inventory I exceeds the sum of the safety stock S and the predicted demand D, the result in parentheses is negative, and after the max operation, 0 is output, indicating that no purchase is needed at present, effectively avoiding over-purchasing and inventory backlog; 6.5 Calculate the hedging ratio H = min(λ × σ / 0.15, 1.0) based on the price volatility σ and the risk appetite coefficient λ, where 1.0 is the upper limit of the hedging ratio, i.e., 100% hedging; the min operation ensures that the hedging ratio H does not exceed 1.0 to prevent excessive hedging; 0.15 is the benchmark volatility reference value. When the actual market volatility σ is equal to 0.15, the hedging ratio H is exactly equal to λ, where λ is the risk appetite coefficient, with a value range of [0,1], which is configured by the company according to its own risk management strategy: the larger the value of λ, the higher the risk appetite and the more willing to hold a higher proportion of hedging positions; the smaller the value of λ, the more conservative the approach and the lower the hedging ratio.

[0024] Furthermore, steps I to VI below describe in detail the training and inference mechanisms of the deep reinforcement learning model, i.e., how the intelligent decision engine learns and continuously optimizes the procurement decision strategies in steps 6.1 to 6.5 above: I. Definition of State Space: S = {P_Futures, P_Spot, V_Trading Volume, I_Inventory, T_Time, H_Historical Price Series (20 dimensions), F_Fundamental Indicators (8 dimensions)}; II. Action Space Definition: A = {Buy Futures, Buy Spot, Mixed Purchase, No Purchase}, each action includes a purchase quantity parameter; III. Reward Function Design: R = -Purchase Cost + α×Inventory Satisfaction - β×Risk Penalty Term, where the risk penalty term includes price slippage loss, position risk, and liquidity risk; IV. Neural Network Structure: An Actor-Critic architecture is adopted. The Actor network outputs the action probability distribution, and the Critic network evaluates the state value. Both networks contain 3 fully connected layers with 256 neurons per layer and ReLU activation function. V. Training process: Using an experience replay mechanism, batch data is sampled from historical transaction data for training, and the network parameters are updated using the Proximal Policy Optimization (PPO) algorithm; VI. Model Updates: The model is incrementally trained weekly using the latest trading data to keep it adaptable to market changes.

[0025] Step 7, as follows Figure 4 As shown, the risk control module performs risk checks on the generated procurement strategy, and the check items include: 7.1 Does the amount of a single transaction exceed the budget limit L1? 7.2. Does the total transaction amount for the day exceed the daily limit L2? 7.3. Does the holding ratio after the purchase exceed the maximum holding ratio R_max? 7.4. Does the current market volatility exceed the abnormal volatility threshold σ_max? 7.5. Whether the available funds meet the margin requirements.

[0026] Step 8: If any risk test fails, the procurement strategy will be rejected, the risk event will be recorded, and an early warning notification will be sent; if all tests pass, proceed to Step 9. Step 9: The automated trading module executes purchase orders. For futures purchases, it sends the order to the futures company via the CTP interface; for spot purchases, it sends the purchase order to the supplier system via the HTTP interface. Step 10: The transaction module monitors the order status in real time and records the transaction price, quantity, and time. Step 11: The data storage module saves a complete record of this procurement, including market data snapshots, decision parameters, transaction results, etc. Step 12: Determine whether to continue monitoring the market. If yes, return to Step 2; otherwise, end the process.

[0027] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and improvements, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. An intelligent futures-spot fusion automated procurement system, characterized in that: The system includes a data acquisition module, a data clarity module, a futures and spot analysis module, an intelligent decision engine, a risk control module, and an automatic trading module. The data acquisition module is configured to collect market data in real time from futures exchanges, spot trading platforms, and industry information websites. The data clarification module is configured to perform outlier detection, missing value imputation, and data standardization on the collected raw data. The futures-spot analysis module is configured to use time series analysis algorithms and statistical models to perform real-time analysis of futures price trends, spot price trends, futures-spot price spreads, and basis changes, and generate market trend forecast reports. The intelligent decision engine is configured to automatically generate procurement recommendations based on a multi-factor quantification model and machine learning algorithms, taking into account price factors, inventory factors, capital factors, and risk factors. The risk control module is configured to set multiple risk thresholds and automatically suspend transactions or send alarm information when a risk warning is triggered. The automated trading module is configured to automatically send trading instructions to futures brokers or spot suppliers through a trading interface based on the instructions of the intelligent decision engine, and to track the order execution status in real time.

2. The intelligent futures-spot fusion automatic procurement system according to claim 1, characterized in that: The system also includes a data storage module, which is configured to use a distributed database to store historical transaction data, market data, and decision records.

3. The intelligent futures-spot fusion automatic procurement system according to claim 1, characterized in that: The intelligent decision-making engine automatically generates procurement recommendations, including: A1. Calculate the futures purchase cost C_futures and spot purchase cost C_spot; A2. Calculate the spot-futures price advantage index K = (C_spot - C_futures) / C_spot × 100%; A3. If K > β, prioritize futures purchases; if K < -β, prioritize spot purchases; if -β ≤ K ≤ β, adopt a mixed purchase strategy, where β is the threshold. A4. Calculate the purchase quantity Q = max(S - I + D, 0) based on the enterprise inventory level I and safety stock S, where D is the predicted demand for the next 30 days. A5. Calculate the hedging ratio H = min(λ × σ / 0.15, 1.0) based on the price volatility σ and the risk preference coefficient λ, where λ is the risk preference coefficient.

4. The intelligent futures-spot fusion automatic procurement system according to claim 1, characterized in that: The risk control module performs risk verification on the generated procurement strategy, including: B1. Does the amount of a single transaction exceed the budget limit L1? B2. Does the total transaction amount for the day exceed the daily limit L2? B 3. Does the holding ratio after the purchase exceed the maximum holding ratio R_max? B4. Does the current market volatility exceed the abnormal volatility threshold σ_max? B5. Do the available funds meet the margin requirements? 5. A method for an automated procurement system based on intelligent futures-spot fusion as described in any one of claims 1-4, characterized in that: The method includes: S1. Collect real-time market data and spot market price data according to the collection frequency, and preprocess the collected data; S2. The futures-spot analysis module calculates key indicators and judges the market status, identifies whether there are arbitrage opportunities or abnormal price fluctuations. If the futures-spot price difference rate exceeds the stock trading threshold α, it is marked as an arbitrage opportunity. S3, the intelligent decision engine generates procurement strategies based on market conditions and enterprise needs, and performs risk verification on the generated procurement strategies through the risk control module; S4. After all risk checks are passed, execute the procurement order, monitor the order status in real time, and finally store the complete record of this procurement.

6. The method for an automated procurement system based on intelligent futures and spot market fusion as described in claim 5, characterized in that: The intelligent decision-making engine generates procurement strategies based on market conditions and enterprise needs, including: A1. Calculate the futures purchase cost C_futures and spot purchase cost C_spot; A2. Calculate the spot-futures price advantage index K = (C_spot - C_futures) / C_spot × 100%; A3. If K > β, prioritize futures purchases; if K < -β, prioritize spot purchases; if -β ≤ K ≤ β, adopt a mixed purchase strategy, where β is the threshold. A4. Calculate the purchase quantity Q = max(S - I + D, 0) based on the enterprise inventory level I and safety stock S, where D is the predicted demand for the next 30 days. A5. Calculate the hedging ratio H = min(λ × σ / 0.15, 1.0) based on the price volatility σ and the risk preference coefficient λ, where λ is the risk preference coefficient.

7. The method for an automated procurement system based on intelligent futures and spot market fusion as described in claim 5, characterized in that: The risk verification of the generated procurement strategy through the risk control module includes: B1. Does the amount of a single transaction exceed the budget limit L1? B2. Does the total transaction amount for the day exceed the daily limit L2? B 3. Does the holding ratio after the purchase exceed the maximum holding ratio R_max? B4. Does the current market volatility exceed the abnormal volatility threshold σ_max? B5. Do the available funds meet the margin requirements?

Images