In all areas of technology, people are talking about machine learning and deep learning, especially the Internet of Things (IoT)market. This week\'s announcement by Google and Salesforce also focused on how to use machine learning and artificial intelligence to improve the way we live, work and play. Obviously machine learning will be part of new products and leading SaaS offerings. As companies seek to build data This feature is very critical. The question is how will companies release data and insights from existing products, and how should business leaders treat new concepts of machine learning? To answer these questions, I spoke with Sara Gardner, chief technology officer at Hitachi insights group. In the world of machine learning, business leaders should follow three core principles, Gardner said. Similar to any IT or business project, IT and OT should start by defining important business questions that can be answered by analyzing large datasets. While this sounds obvious, many companies are trying to choose an important business goal and define the scope of the goal to achieve it. For example, the direction we need to reduce costs is not clear enough. You have to box the problem into a problem that data can solve. For example, a mining company may want to reduce the number of days of downtime and maintenance costs associated with the failure of truck equipment. Considering that a pair of truck tires for a mine car can cost hundreds of billions of dollars, a good start is to find problems that cause tire failures. Machine learning for predictive maintenance can save money and minimize downtime. The next step is to define the data available to the organization. Many companies have collected a lot of information, but have not yet used these resources. Others will have to use instruments and equipment with sensors to collect data. Machine learning needs to be learned from experience, and in the case of certain devices, there is limited data on failures that you may need to analyze. In this case, you need to create the fault using the simulator. Taking into account the amount of data that organizations can access, Gardner notes that it is important to identify key metrics around the issues you are trying to solve. For example, the change in tire pressure will be a key indicator in the mining use case-machine learning techniques are also helpful here. As part of step 2, you\'ll also need to look for tools that can absorb a lot of data and help you visualize important trends in that data. Gardner noted that this is one of the reasons why Hitachi acquired Pentaho, which provides big data tools to extract, prepare and mix structured, unstructured and semi-structured data Unstructured data It also provides visualization and analysis. Given the discussion with enterprise customers at Lopez Research, I recognize that companies can do this using multiple types of tools. The challenge is to choose the right tool for the right job. Vendors like Hitachi are also aware of the importance of including such products in their kits. Now we are taking the hardest step, the third. Your company understands the goal, you have the data, but someone has to define a set of methods for analyzing the data. Your organization must address issues such as how to propose a standard model. The technical team must define what kind of algorithm makes the most sense for the problem you are trying to solve and the type of data you have. Understanding these methods is where ever elusive data scientists work. I learned from discussions with many vendors and end-user IT stores that data science is part of art, part of science. Exploration and trial and error are part of the process of creating algorithms. Machine learning means that you are building a system that constantly learns and adjusts algorithms based on what you have learned. This type of analysis is very different from the batch data processing and weekly business intelligence dashboards used by many companies today. But what if you don\'t have a stable data scientist yet? To fill the gaps in data scientists, companies must look for suppliers. I note that more and more companies are trying to address the lack of data scientists by providing customers with a set of ready-made algorithms or solutions. As far as Hitachi is concerned, it is developing a blueprint and a solution core for its major industries. These products range from machine learning algorithms to hardware that processes data. The solution core should provide more than 60 to 80% of the required functionality. A company can add these solutions through its employees or work with a service company to complete them. Sales staff and other personnel. Com is embedding the data analysis process in applications such as its CRM and service cloud. At the same time, Oracle has a series of IoT PaaS Services integrated with existing business applications. Since there is no one-size-fits-all approach to data analysis, companies should combine these solutions to meet the analytical needs of various parts of the business. The last part of the strategy is to define how you use insight. Professional knowledge in the field is as important as mathematics. Without domain expertise, it is difficult to apply data trends in ways that maximize business value. For example, it is costly to take a device out of the service for maintenance. There are usually several options for how to deal with this problem. You can reduce the time the device is running, and you can also increase the frequency of maintaining the device to reduce downtime. You can extend the maintenance window if you have an early warning system. Conclusion: machine learning tools will provide the potential to unlock new insights and drive business value. Machine learning, however, is not a solution where you can simply purchase and go directly into your organization for insight. It needs to experiment and coordinate between business and It to improve efficiency and competitive advantage.