A few days ago, two business decision optimization solver software independently developed by China successfully topped the international authoritative mathematics decision software evaluation rankings. Shanshu Technology took the top spot, followed by Ali, which aroused the attention of the Chinese people to decision optimization solvers. . Previously, due to international competition, chips and operating systems have become a "stuck neck" project that China wants to make a national breakthrough, and business mathematics and scientific simulation computing software such as Matlab have been disabled, making business decision optimization solver software a chip and After operating the operating system, it is necessary to master the country's important weapon of independent intellectual property rights.
The business decision optimization solver software is mainly in the form of engineering software, using mathematical optimization algorithms to optimize and solve large-scale and complex problems in public institutions and commerce, such as transmission network planning and generating unit combination optimization scheduling, logistics distribution route planning and optimization, production Manufacturing scheduling, commercial store location planning and optimization, etc., have extremely important value and significance in actual production and life. When the actual problem is more complex and the scale of the problem is larger, the more reliable large-scale optimization tools are needed to provide quality assurance and enhance decision-making confidence.
The internationally renowned large-scale commercial optimization solver software Gurobi has been used in China’s petroleum and petrochemical, steel metallurgy, logistics and express delivery, e-commerce warehousing, manufacturing, financial investment, media management, network communications, air transportation, power markets and other industries. It is widely used. As of the end of 2019, the cumulative number of applications for academic licenses in China has exceeded 50,000. Another well-known solver software, IBM ILOG CPLEX, also uses data science and mathematical algorithms for the optimization of large-scale business problems, such as helping a logistics company redistribute railway transportation routes and save $26 million a year.
Professor Hans Mittelmann of Arizona State University in the United States has evaluated a variety of open source and commercial mathematical programming solvers for nearly 20 years, and is recognized as a reliable third-party evaluation platform. At the end of August 2020, the solvers developed by Shanshu Technology, Alibaba Dharma Academy and Gurobi ranked the top three in the linear programming simplex method evaluation, and two domestic ones were included. Among them, Shanshu Technology has been on the list since its inception in July 2019. Now that Gurobi has re-entered the evaluation, it also demonstrates the world's top level of Shanshu solver's linear programming capabilities.
The Great Power Controversy: Business Decision Optimization Solver
Why is it said that business decision optimization solver software is as important as the chip and operating system? Or in other words, why is the independent research and development and localization of commercial decision-making optimization solvers a great power struggle? What is the significant impact of the business decision optimization solver on the national economy and people's livelihood? First look at an example.
The global grain supply is a huge system. Taking France as an example alone, about 70 million tons of grains are produced every year and the various distances transported to the world add up to 1 billion kilometers. Warehousing suppliers are an important part of the global grain supply chain. They collect and transport grains to warehouses when the grains are ripe, and then distribute them to customers throughout the following year. Previously, grain warehousers chose transportation routes and warehouse locations based on experience. A typical grain warehouser had to manage hundreds of different grains and hundreds of different collection and storage locations, as well as a large downstream customer base. The combination is extremely large. In addition, grain warehousers also face risks and uncertainties brought about by weather. It often takes until the last minute to know the quality and quantity of the collected grains, and then formulate corresponding storage and sales strategies.
A French software company that aims to solve the problem of agricultural product collection, transportation and storage optimization, uses agricultural supply chain big data and mathematical optimization algorithms to produce optimized logistics solutions. Through integrated business decision optimization solver software, it can bring A 10-15% supply chain cost saving, taking the agricultural product supply chain in France as an example, this means a saving of 1 billion kilometers per year, which is equivalent to a cost saving of 3 billion euros per year.
Decision-making optimization solutions can bring annual cost savings of 3 billion euros for French crop logistics. This is undoubtedly a national importance for China as a major agricultural country, a major logistics country, and a major food consumer, especially for China in supply-side reforms. The strategic value of the device.
The Chinese scene creates the next generation of world-class software
The three giants of American business decision solvers Gurobi, CPLEX, and CPLEX in Xpress have been born for more than 30 years and were later acquired by IBM. The founder of CPLEX later founded Gurobi in 2008, and the American company Fair Isaac acquired the British company Dash Optimization in 2008, which is the founding company of Xpress software. Since 2008, the business decision solver industry in the United States has begun to develop, and it has been refined and iteratively evolved according to the scenarios of American companies and other countries around the world. In addition to the Big Three, Matlab, SAS, PTC and other companies also provide corresponding solver functions in their industrial engineering and scientific computing software.
So, whether it is Shanshu Technology or Alibaba, why have defeated mature American software in the most important business decision optimization algorithm field in recent years? First of all, the field in which Shanshu Technology and Ali compete is called simplex linear programming. In the Mittelmann list, there are other linear programming algorithms and other mathematical optimization algorithms such as integer programming and nonlinear programming. The simplex method of linear programming is considered to be the beginning of modern mathematical programming and also the beginning of operations research. It is one of the top ten algorithms of computational science in the 20th century. Mathematical programming, also known as mathematical optimization, is an important branch of operations research, including linear programming, integer programming, nonlinear programming and other research directions, and the linear programming simplex method is one of the important algorithms.
Dr. Ge Dongdong, co-founder of Shanshu Technology and project leader of COPT Shanshu optimization solver, graduated from the Department of Management Science and Engineering of Stanford University. He is currently a director of the Chinese Society of Operations Research and the dean of the Interdisciplinary Research Institute of Shanghai University of Finance and Economics. He has hosted many countries And provincial scientific research projects, published many academic papers in top international journals and conferences, and provided technical services for many domestic and foreign benchmarking enterprises. Ge Dongdong introduced that business decision-making optimization solver software based on mathematical optimization algorithms is mainly to solve practical software engineering problems. Real business scenarios are needed to polish excellent solver software. In the past few years of traditional Shanshu solver development, the domestically provided industrial scenarios used for testing have reached hundreds of millions of system solutions, providing extremely rich resources for the testing and improvement of solver software.
Since 2015, my country's economy has entered a new stage, and the original economic growth momentum is no longer sufficient to support new development needs. In November 2015, the central government first proposed the "supply-side reform", emphasizing the continuous optimization of the supply-side and generally increasing social productivity. In 2016, China’s big data industry and the penetration of big data technology into government and enterprise have reached a staged result. The government and enterprise have completed the first wave of big data accumulation, and there is an urgent need for advanced analysis technology to accumulate big data. For analysis and production optimization. All these objectively optimize the solver software for business decisions and open up the market space. On the other hand, after more than 40 years of development in reform and opening up, China has become the world's largest manufacturing country, online retailing country, trade largest country, and world's second largest economy. The scale of China's business scene has far surpassed that of Europe and America. area.
Take the production scheduling optimization project provided by Shanshu Technology for a large domestic private manufacturing company as an example: the manufacturing company has hundreds of factories and hundreds of workshops, and each factory has dozens of suppliers, involving tens of thousands of varieties. Raw materials, parts and secondary components, etc., the constraints include environmental protection requirements, export requirements, special supply material distribution and other complex conditions. It is necessary to put all factories together for production scheduling optimization-at the beginning of each month, the next 30 days For precise production scheduling, it is required to refine the production plan for each workshop every hour, and then make a rolling forecast for the following 26 weeks-the variables involved are as high as 50 million to hundreds of millions. In the past, this manufacturing company, which is already far ahead in domestic digitization and refinement, used American production scheduling software to specify the optimal plan for each factory. After each factory optimizes the production schedule, it also requires a lot of manual intervention to coordinate production. Now, we have adopted the idea of putting all factories together for overall optimization. Although we can better do overall optimization, the problem planning is huge. The manufacturing company even invited some professors from MIT, but in the end Shanshu Technology realized the "30+26" large-cycle production scheduling optimization within 3 hours, and the daily optimization only took less than 2 hours, which was successfully achieved. A localized alternative that is better than foreign solvers and foreign solutions.
Business decision optimization with solver as the core
The business decision optimization solver, like the chip and operating system, cannot solve the actual business decision optimization problem alone. Instead, the actual business decision optimization problem needs to be mathematically modeled first, and then the solver is used to solve it. According to Luo Xiaoqu, co-founder and CEO of Shanshu Technology, Shanshu Technology is currently the only independent business decision optimization solver software provider in China. It also provides modeling services for complex business decision optimization problems, using operations research and mathematical optimization algorithms and manual Intelligent algorithms, etc., provide enterprises with large-scale business decision-making optimization solutions, and the company's business has entered an explosive period in 2019.
In addition to the large-scale production scheduling optimization problem of a manufacturing company mentioned earlier, Shanshu Technology has also optimized the location of a retail chain supermarket in China. The location selection method adopted by the retail chain supermarket before is to divide the city with an area of 200 meters and 200 meters, and then send people to sweep the street for data statistics. The statistics include whether there are green belts around the commercial real estate, whether the steps exceed 3 levels, and whether the surrounding areas There are KFC and ATM, surrounding population information, and so on. Negotiations after entering the screening range are too complicated, labor-intensive, and costly. Shanshu Technology takes the maximization of expected revenue as the optimization goal, and performs optimization analysis and calculations based on various big data. It gives a score of "0 or 1" to the small grid of 200 meters and 200 meters. Compared with the work results, if it can cover 70% of the previous results, it is considered effective, and Shanshu Technology has actually reached 80%-90% coverage and is more accurate. The retail chain supermarket finally chose the solution of Shanshu Technology, which can cover the same population with 40% of the similar snack chain stores in Shanghai alone. It not only saves the manpower and cost of sweeping the street, but also improves the efficiency of store location selection.
Luo Xiaoqu said that when Shanshu Technology was established in 2016, the introduction of the concept of business decision optimization to the company did not arouse high recognition from the company. In the past two years, enterprise big data construction has entered a new stage. The original reliance on reports and large-screen visualization can no longer reflect the results of big data construction. A large number of leading companies have begun to focus on big data and use advanced Analysis and mathematical optimization algorithms optimize and analyze actual business decision problems. Using data to drive decision-making and data analysis to speak has become the new normal of business operations.
Especially my country has developed intelligent manufacturing in recent years. A batch of lighthouse factories have emerged. Intelligent manufacturing has moved from the construction of intelligent production lines and equipment on the cloud to the use of advanced algorithms for production scheduling, production planning, and production, supply and marketing coordination. stage. As a large consumer country, my country has seen consumption upgrades in recent years. Consumers’ demand for different consumer products has increased dramatically. This has led to strong uncertainty. The ever-changing consumer demand has led to highly demanding varieties and small batches. With flexibility in the production supply chain, many leading companies have entered the stage where large-scale optimization of business decisions is required.
Ge Dongdong emphasized that ranking on the international authoritative solver software list does not mean that it can really solve the actual large-scale business decision optimization problem. The ranking itself only requires small skills in algorithm optimization, and actual business decisions. The problem is the real challenge. The core product of Shanshu Technology is the COPT optimization solver. It currently provides open source, stand-alone and server versions, including linear, mixed linear integer programming, nonlinear optimization and other modules. Since the release of COPT 1.0 version that can solve large-scale and complex business problems in July 2019, the solver has successfully dealt with the simplex linear programming problem of 50 million to hundreds of millions of variables, and its integer programming module has also been successfully applied Domestically produced alternatives for sensitive projects in many countries.
As a domestic software, Cedar COPT solver currently supports all mainstream operating systems (all 64-bit systems) including Windows, Linux and MacOS, and provides Python, PuLP, Pyomo, C, C++, C#, Java, AMPL Interface with mainstream computing languages such as GAMS and support ARM64 platform. This solver has been widely deployed in Shanshu Technology’s smart supply chain and Industry 4.0 smart systems in the past two years. It has been used in a large number of leading companies, including Budweiser, Orion, Xiaomi, COMAC, etc., and is the only national software manufacturer. It has made unique contributions to many major national projects such as deep space exploration, civil aviation, power grids, petroleum, and information security. Unlike Shanshu Technology, Alibaba mainly provides solver services through Alibaba Cloud.
In summary: 2020 is the year of the opening plan for the 14th Five-Year Plan. It will also coincide with the strategic overall situation of the great rejuvenation of the Chinese nation and a century of unprecedented changes in the world. The business decision optimization solver is as important as the chip and operating system. It plays an extremely important role in the large-scale optimization of public services and commercial operations. Chinese technology companies represented by Shanshu Technology and Ali have taken the lead. , Make unremitting efforts to build hard technology competitiveness in China! (Text/Ningchuan)