The CRISP-DM(CRoss Industry Standard Processfor Data Mining) model is a methodology developed in 1996 by about five large tech companies. With the goal of standardizing data mining activities. The CRISP-DM methodology is still one of the most widely used analysis models for a data coach, data scientist and a data analyst. The model is very dynamically designed with the ability to navigate between different process steps. In practice, it is often combined with other Lean and Agile project management approaches.
The first stage is immediately an important one. This involves really understanding the cases. During this phase, preparatory work is done by creating a problem definition and a strategic plan for your objective. In order not to have to make a radical change in direction at a later stage, it is crucial to know what resources you want to work with to achieve your objective.
This phase involves diving deeper into the data. Take a good look at what data is available and whether it contributes to achieving your goals. Identifying which columns contain incorrect or missing data is of great importance because these can give a clouded view of the situation. Framing metrics is therefore a logical next step, when working in a large team there should be no difference of interpretation.
Now that you have a good understanding of the available data, the next stage is to detect columns that contain the same data but are displayed differently. For example, a column might include a person's height but it is represented in three different units of measurement cm, m and foot. During this phase, the column is cleaned up so that there are no misunderstandings and all data meets the same requirements.
In this phase, the goal is to model and transform the data so that it is ready to be visualized. Often one wants to be able to observe a particular trend within a specific group or classification.
In this example, a web shop wants to measure the influence of visit duration on, for example, buying behavior. Based on this, groups of users are clustered per three minutes they spend on the website, then it is examined whether there is any correlation between the average number of conversions achieved by visitors from these groups.
These are the stages that test whether the discoveries meet the demanded criteria. This is where we look back at the entire process and how these statistics were arrived at. Here we check one more time that there are no errors in the data preperation and modeling. When all these factors have been validated, we can proceed to the final step, Deployment.
The last phase that one enters is the deployment phase, in which it must be ensured that the insights obtained are displayed in a clear manner. So that in the future data can be collected and tracked in a structured way.
Is your organization running into issues regarding the collection, storage, validation or modeling of data. Our data coach is happy to take a look with you without any obligation. Request a meeting here.
By collaborating with educational institutions and industry combined with technology to accelerate digitalization, we are developing knowledge on "what, why, how" of the digital factory together with our members.
Dutch