{"id":22882,"date":"2020-03-19T17:12:52","date_gmt":"2020-03-19T16:12:52","guid":{"rendered":"https:\/\/bes-h.com\/es\/?p=22882"},"modified":"2020-04-15T17:03:43","modified_gmt":"2020-04-15T15:03:43","slug":"modelos-parametricos-y-no-parametricos-dentro-de-la-estructura-big-data","status":"publish","type":"post","link":"https:\/\/bes-h.com\/en\/modelos-parametricos-y-no-parametricos-dentro-de-la-estructura-big-data\/","title":{"rendered":"The parametric and non-parametric modelling for Big Data"},"content":{"rendered":"

The Data Science team from Stratesys, in particular the subject matter expert Yaroslav Hernandez Poitomkin<\/a> explains what is Big Data, how it is applied to different business departments and what are the differences between the parametric and non-parametric modelling.<\/strong><\/em><\/p>\n

For further information, you can get in contact with the experts<\/a>.<\/p>\n

 <\/p>\n

Nowadays we are experimenting a data revolution, which means more capacity for data storage<\/strong> and more data volume<\/strong>.<\/p>\n

Also, more variety due to newly created tools and ways of communication such as social media, webs and apps. And finally, data velocity<\/strong> and data processing related to data generation processes and business models constraints.<\/p>\n

These three concepts, Volume, Variety and Velocity define the \u201cBig data\u201d term<\/strong>. Therefore, if this definition can be applied and integrated into your business processes, it means that you might have the needs of massive data processing and\/or analysis.<\/strong><\/p>\n

 <\/p>\n

Many organizations believe that advanced analytics can potentially enhance their business processes related to logistics, marketing, operations and strategy<\/strong>.<\/p>\n

Often, these processes are covered by data modelling techniques and methodologies<\/strong> as the core engine of the decision support systems. The decision support systems allow discovering valuable knowledge from the data that reflects the business status and trends<\/strong> always taking into account historical information.<\/p>\n

Therefore, an assisted decision making is much more effective than just expert knowledge and intuition only.<\/p>\n

 <\/p>\n

 <\/p>\n

Difference between the parametric and non-parametric modelling.<\/h4>\n

Now, we can introduce a more technical part. There are many tools and solutions related to \u201cBig data\u201d and advanced analytics. The real-world problems are very complex and it is very important to choose<\/strong> an appropriate technique or it even might be the case in which one would need to design their own custom methodology<\/strong>.<\/p>\n

There exist two main types of models: parametric<\/em> models and non-parametric<\/em><\/strong> models.<\/p>\n

In a parametric model<\/strong>, the data follows a probability distribution with a fixed set of parameters. This is a very appealing setting because the initial assumptions of the model<\/strong> are fulfilled<\/strong> and it is well-based from the theoretical point of view<\/strong>.<\/p>\n

The problem with it might arise, for example, when one has multimodality behaviour in the data, which means that many models such as Quadratic Discriminant Analysis (QDA) or Linear Discriminant Analysis (LDA) (Hastie et al. 2008) are not suitable anymore and that the inferences one does might be completely misleading. But for instance, Fisher Linear Discriminant Analysis (FLDA) (Hastie et al. 2008) doesn’t need the normality assumption, neither Principal Component Analysis (PCA) (Jolliffe, I.T. 2013).<\/p>\n

 <\/p>\n

\u00a0The non-parametric models<\/strong> can be regarded as:<\/p>\n