These are a few steps necessary so that the data science journey to successfully complete a data science program:
Step 1: Diagnostic
This is a crucial step where we help to identify the problem and how Machine Learning may help to leverage your business.
- What is the problem your are trying to solve?
- What resources and data are / can be available? Is it clean, reliable, easy to access/migrate?
- Is your data available?
- Are your goals feasible?
- What legal and regulatory concerns are important?
Step 2: Strategy
A good Data Science strategy This is cornerstone for the success of any DS project.
- What is a DS strategy and how to be implemented?
- How do you maximise the value from the data you have?
- How your data science strategy aligns with company strategy
Step 3: Resources
- Do you have resources? How realistic are they to achieve the goals?
- HR: What autonomy do they have?
- HR: How are they managed?
- What physical resources do you have and can use?
Step 4: Execution
How is the roadmap to execution? I will explain why it is important to balance quick wins with long term goals. How to set the right KPIs, manage the team and setup the conditions for perfect execution.
Step 5: Culture
Artificial Intelligence and Data Science is becoming prevalent in every business. However, not all organizations are prepared for Data Science. If the company doesn’t have the right culture ready to accommodate the transformations imposed by the Data Science team, then it may be a waste of resources. Successful data science projects are painful and much more than hiring clever data scientists.
What are the key phases and outputs of an AI project?
Project Planning and Preparation
| Data Access | The department must ensure that the data required for the project is accessible. | SMEs | Access to the data |
| Knowledge Transfer | SMEs should provide data documentation and other business process documentation required to ensure that the project can be delivered | SMEs | Scientists have an overview of the data content and business requirements |
Data Science Project Execution
| Scientific Design | Data scientists will decide on methods and techniques required to address the problem | Lidinwise SMEs | Implementation Plan |
| Exploratory Analysis | Exploration of data and summary (this includes cleaning, data preparation and feature engineering) | Lidinwise | Exploratory Analysis Briefing |
| Train/Test (offline) | train -> validation -> test | Lidinwise | Model results (offline) |
| Implementation and deployment test | This will be an iterative process and will likely involve online test(s) | Lidinwise SMEs | Pipeline test and Model results (online) |
Project Close out
| Full production | Full implementation of model into current processes; Training and knowledge transfer to Department | Partnerships | Final ML production pipeline |
| Documentation | Write up description | Lidinwise with participation | Document with a description of the project delivered |