登录查看更多内容

How do you prevent scope creep in machine learning projects?

由人工智能和领英社区提供技术支持

Scope creep is the tendency of a project to expand beyond its original goals, often due to changing requirements, unclear expectations, or lack of communication. It can lead to wasted time, resources, and quality in any kind of project, but especially in machine learning, where data, models, and algorithms are complex and dynamic. How can you prevent scope creep in your machine learning projects and deliver value to your stakeholders? Here are some tips to help you.

此文章中的业界达人

由社区从 5 条内容中精选。了解更多

1 Define the problem and the success criteria

Before you start any machine learning project, you need to have a clear and specific definition of the problem you are trying to solve and the metrics you will use to measure the success of your solution. This will help you avoid getting distracted by irrelevant or unnecessary features, data sources, or methods that do not contribute to your goal. You should also align your problem definition and success criteria with your stakeholders, such as clients, managers, or users, and get their feedback and approval. This will help you avoid misunderstandings or conflicts later on.

添加您的观点

Younes Sandi

Data Scientist | AI Consultant | Mentor | Recognised Talent in Tech by UK Global Talent Program ????
举报内容
Success or definition of done might mean different things from different perspectives, for example a data lead might define success in reaching a specific ML performance, let’s say minimum 0.75 and 0.8 in precision and recall respectively if it’s a supervise classification problem while a product manager measures that in terms of business value, let’s say they care about the model uplift and customer retention increase by 15%. It is crucial to understand these requirements and the trade of between them, as most of the time it’s impossible to achieve all these objectives considering project limitations.

已翻译

赞

2 Plan your project scope and timeline

Once you have defined the problem and the success criteria, you need to plan your project scope and timeline. This means breaking down your project into manageable tasks, estimating the time and resources needed for each task, and setting milestones and deadlines. You should also prioritize the tasks according to their importance and urgency, and identify the dependencies and risks involved. Planning your project scope and timeline will help you stay focused and organized, and track your progress and performance.

添加您的观点

3 Communicate and document your project

Communication and documentation are essential for preventing scope creep in machine learning projects. You need to communicate regularly and clearly with your team members, stakeholders, and other collaborators, and update them on your project status, achievements, challenges, and changes. You also need to document your project process, decisions, assumptions, results, and feedback, and store them in a central and accessible location. Communication and documentation will help you avoid confusion, errors, duplication, or misalignment in your project, and ensure transparency and accountability.

添加您的观点

4 Manage change requests and expectations

Despite your best efforts, you may encounter change requests or expectations from your stakeholders or yourself during your machine learning project. These may be due to new data, insights, requirements, or opportunities that arise along the way. While some changes may be beneficial or necessary for your project, others may be unrealistic or detrimental. You need to manage these change requests and expectations carefully and critically, and evaluate their impact on your project scope, timeline, quality, and value. You should also communicate the pros and cons of the changes to your stakeholders or yourself, and negotiate a mutually acceptable solution.

添加您的观点

Younes Sandi

Data Scientist | AI Consultant | Mentor | Recognised Talent in Tech by UK Global Talent Program ????
举报内容
When decided to add a change request it’s a good practice to communicate with stakeholders to understand the level of effort (time, volume, resources, etc), complexity of the task, business value and priority. This way you’ll have a much better idea how to approach the change request.

已翻译

赞

5 Review and evaluate your project

Finally, you need to review and evaluate your project regularly and systematically, and compare it with your original problem definition, success criteria, project scope, and timeline. This will help you identify any deviations, gaps, or issues in your project, and take corrective actions if needed. You should also solicit feedback from your stakeholders or yourself, and assess how well your project meets their or your needs and expectations. Reviewing and evaluating your project will help you ensure quality and consistency, and deliver value to your stakeholders or yourself.

添加您的观点

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?

添加您的观点

Rahul Zende

Data Science at Navy Federal Credit Union | UW Seattle MSIM | Mentor
举报内容
Scope creep isn’t new, it is often observed outside of ML projects too; here are some tips to combat it. - A robust requirements & scoping team/framework that helps defend the ML teams against additional asks can play the first line of defense to confirm whether those are truly urgent and/or important. - Leaders managing ML teams need to be experts in dynamically categorizing additional asks into buckets like: a.trivial, b.non-trivial but doable, or c.downright unrealistic; this can ensure best bang for buck in agile scenarios. - Pipelining stakeholder asks is a good practice. They need to be coerced into a system that splits demands into multiple iterations consisting: minimum viable product goals, stretch goals, and north-star visions.

已翻译

赞
Sarah Floris, MS

Senior Data & ML Engineer | dutchengineer.substack.com | Host of Ask A Data Mentor Podcast
举报内容
I highly recommend reading the first part of Machine Learning in Action by Ben Wilson; he covers the pitfalls of most projects and how to prevent them.

已翻译

赞
Manishankar Singh

Machine Learning Engineer at Tata Consultancy Services
举报内容
The best way to prevent scope creep is to have requirement capturing done in a transparent way and available for everyone in team to check in a digital format. Additional discipline can be brought in by enforcing input and acceptance criteria for each additional work. Most of the time scope creep happens because the person giving the requirements does not have substantial depth in ML/data analytics and might be believing in hype without ever doing hands on.

已翻译

赞

Machine Learning

+ 关注

给文章评分

我们借助人工智能创建了此文章。您认为这篇文章怎么样？

很棒不太好

举报此文章

查看全部

How do you prevent scope creep in machine learning projects?

1

2

3

4

5

6

1 Define the problem and the success criteria

2 Plan your project scope and timeline

3 Communicate and document your project

4 Manage change requests and expectations

5 Review and evaluate your project

6 Here’s what else to consider

Machine Learning

给文章评分

感谢您的反馈

更多Machine Learning相关文章

更多相关阅读内容

How do you prevent scope creep in machine learning projects?

1

2

3

4

5

6

1 Define the problem and the success criteria

2 Plan your project scope and timeline

3 Communicate and document your project

4 Manage change requests and expectations

5 Review and evaluate your project

6 Here’s what else to consider

Machine Learning

给文章评分

感谢您的反馈

查看其他技能