Leveraging machine learning is a solid step toward ongoing, impactful data analytics. Below, 10
Forbes Technology Council members share the first steps for any tech department that wants to start improving their data analytics through machine learning.
1. Find the right problem.
Do not try to boil the ocean. Your first ML project should just be a pot of water. The temptation is to solve a really big problem—scale it back. There are always smaller issues within. Pick one that has plenty of data behind it. Your odds for success are higher with smaller issues that have a lot of data as opposed to big items with sporadic data. -
David Moise,
Decide Consulting2. Develop a business plan and use cases.
It would be best to get people interested and excited about the possibilities—with a touch of urgency. Develop a business plan and use cases. Let people know that the process will not be simple and it will involve a commitment of time and resources from a supportive organization to sustain the initiative. The results will make the investment in time and talent well worth the cost. -
Will Conaway,
The HCI Group/Tech Mahindra3. Create a strategy aligned with business goals.
The first step to using ML to improve data analytics isn’t technical at all—it’s strategic. You have to create an ML strategy that is aligned with your business goals and KPIs. For example, if your goal is to get on top of the Google SERP and your KPI is increasing your website’s domain authority, you could build a strategy for using ML to optimize your internal link profile. -
Gergo Vari,
Lensa, Inc.4. Ensure you’re ready to leverage ML for analytics.
Machine learning can be enormously powerful for data analytics, yet ML models are only as good as the data and people who create them. To leverage ML for data analytics, first ask if you have clearly defined problem(s) that are well suited for ML. Ensure you have high-quality data—labeled or self-supervised, depending on the problem—to train the machines and the appropriate ML talent to get the job done. -
Igor Jablokov,
Pryon5. Identify the data that addresses your questions.
Companies looking to incorporate machine learning in their analytics process must first delineate the question of interest, find fit-for-purpose data and leverage appropriate technology to obtain credible results quickly. In the healthcare field, there is a plethora of data available (e.g., from electronic health records), which holds great promise for innovations in patient care. -
Joe Menzin,
Panalgo6. Automate data gathering systems.
Actionable ML insights can change the course of a business, but they’re often only possible if the data quality supports learning the right correlations. Firms should begin by investing in the automation of ingestion, normalization and deduplication of heterogeneous data. Off-the-shelf models powered by great data will almost certainly yield better outcomes than world-class models acting on messy data. -
Andrew Sellers,
QOMPLX, Inc.7. Improve the quality of your data.
The learning on good data is accurate. The first step in improving data analytics is to improve the quality of data. High-quality data is relevant to business, marked with the associated context(s) and a structure that is automation-friendly. -
Vipin Jain,
Pensando Systems8. Clean up and standardize your data.
The first and most important step for a company planning to integrate ML across their analytics function is to standardize and cleanse their data. This step often gets overlooked, but it’s important to ensure any biases or inaccuracies in the data aren’t reflected in the ML results. Clean, standardized data equals more valid and reliable ML outcomes. -
Chetan Mathur,
Next Pathway9. Audit and organize your data.
Audit available data and organize it in a way that allows you to access it consistently and at scale. For humans working on ML projects, it’s important to keep in mind the “garbage-in/garbage-out” principle. The ultimate task of any ML is to learn patterns from data, so if you enter incorrect data, you can expect to produce flawed interpretations, which could cost your organization if the issue is not identified. -
Chris Paquette,
DeepIntent10. Remove data ownership silos.
Data drives machine learning. You must be open to allowing or enabling the data scientist to see and use the data. What we need to focus on is removing the silos of ownership and enabling the data to work. The first step is to be transparent and discuss the benefits and the “how,” not focus on who has the data and where it is. -
Gene Yoo,
Resecurity, Inc.
Không có nhận xét nào:
Đăng nhận xét
Lưu ý: Chỉ thành viên của blog này mới được đăng nhận xét.