Expert tools would help the Phisician to take decision- The process ought to be ethical, feasible and accepted by a new era of hinghly technological phisicians. A new era
INFERENCE AND SUSTAINABILITY IN HEALTHCARE. Kniowledge Discovery from data warehoused in Database KDD
For scientific and economic reasons, beyond data mining we need to discover the knowledge extracted from healthcare data that resides in databases (KDD), these consideration oght to be pllied also to any other knowledge arena.
The information grows exponentially, recently we have assisted to an international forum of innovation in which we hear that now, in Spain, in just only two days it is generated the same information that was generated in the complete year 2003; about 5 exabytes or what is the same; 5 times 10 ^18 bytes, and this is possible because the processing power and storage are cheap currently.
Data ownership is highly important, so far than that, nowadays, in the US, organizations value are based on the data the companies own.
In hospitals there are many data (images and proper data) stored in medical records, these are historical and static data, but we watch an increasing trend; systems are receiving data or images from a mobile pone or any other sensors dinamically, those data are being taken up from patients by devices that measure parameters of a wide and varied nature.
With respect of the knowledge from data (aisled data could not contain relevant information for human being), we have to debug them, treat them with a methodology and then transform, consolidate and merge the various informations sources into knowledge. When we group structure and interpret data, we obtain humanly relevant information that we can use to make decisions, reduce uncertainty and make calculations.
For any field of knowledge, the traditional method of converting data into knowledge resides in manual analysis and expert interpretation. Still in the specific case of the healthcare industry, regularly, it is common in specialities to analyze current trends and changes in health data quarterly by phisicians. After this analysis, the experts issue a detailed analysis report for the sponsors; This report becomes the basis for future decisions and planning in healthcare management arena.
And this traditional process is true in any field of knowledge. But this form of knowledge generation is slow, costly and extremely subjective. In fact, this type of manual analysis is not practical in many domains and we sincerely think that is unsustainable.
Therefore, as a knowledge-based society, is an economic and scientific need for us to move towards a system capable of extracting knowledge from data warehoused in a database (Knowledge Discovery in Database or KDD) and this relates to an abstract level development of methods and techniques to make sense of the data and bring knowledge to health professionals and to the whole industry, thus we are talking about that it is esential to modify the order of the planning and decision-making is taking place in health management, objectifying it. Firstly we need to provide the knowledge from data to the expert and to the manager, and after, these sector specialist will make decision and plan healthcare management base on the objective knowledge they own.
Our challenge then, is to transform the low-level data into more compact outcome such as a small report, or something more abstract ( for example, a model of the process that generated the knowledge), or also more useful, such a predictive model for estimating the value of future cases.
We are taking into consideration that low level data are too voluminous for human beings to be easily understandable and digestible, for example, in a medical diagnostic application we can find data millions and thousands of fields of information, the question to be asked; How can we digest these data and fields manually?.
In the core of the process is located the application of data mining methods to discover patterns and extract knowledge, but In healthcare as in any other knowlegde área, in the process of transforming data into knowledge, we interact with healthcare professionals, which validate our findings; methods and patterns. Each release of our work, thanks to the feedback of the phisician or the expert, is getting progresivelly much more accuracy to our approach towards a solution for healthcare problems.
The KDD is ultimately an attempt to address the problem of the digital information age in which we live; an information overload.
KDD was coined in 1989 to emphasize that knowledge is the final outcome of a discovery obtained from the data, and has been popularized in the fields of artificial intelligence, machine learning or machine learning.
The famous data mining is just only one step in the overall process that comprises the KDD and consists in the application of specific algorithms to extract patterns from the data, but the KDD is incomplete and will be invalid without first preparing the data, select the data that interest us (Target data), clean these data, incorporate the appropriate prior knowledge, and interpret data mining. All these steps are necessary to assure us that useful knowledge is derived solely from the data. The blind application of the single step of mining methods can be dangerous because we can easily lead the data nonsense and the usage of invalid patterns.
The overall process requires an interdisciplinary course of action that evolves continuously from the intersection of research fields such as: the machine learning, pattern recognition, databases, statistics, artificial intelligence (AI), the acquisition of knowledge expert systems, data visualization and high-performance computing (efficient processing of large data volumes for example in mobile environment).
Data mining is mainly related to machine learning, pattern recognition, and statistics to find patterns from the data, while the KDD emphasizes the focus on global knowledge discovery process, and also includes, how are the data stored and how to access to them, how can we scaled-up algorithms for massive data sets and still being able to process them efficiently, and how can these data be interpreted and displayed, how can we model and support those data, and finally how to dessign an useful overall interaction between man and machine.
The whole process should be seen as a multidisciplinary activity that encompasses techniques beyond the scope of any particular discipline, such as machine learning. In this context, there are clear opportunities for other fields of Artificial Intelligence (besides machine learning) to contribute to KDD.
The KDD places a special emphasis on finding understandable patterns that can be interpreted as useful or interesting knowledge. For example, neural networks, although powerful modeling tool, are (under the umbrella of artificial intelligence) relatively difficult to understand compared to decision trees.
Research fields related to Artificial Intelligence (AI) include machine discovery, which focuses on the discovery from observation and experimentation to discover empirical laws, and causal inference for causal models in particular statistical modeling, it has much in common with KDD.
The KDD is fundamentally a statistical endeavour. Statistics provides a language and framework for quantifying uncertainty which is essential when trying to infer general patterns from a particular one, using a sample of a population.
KDD can also be seen as encompassing a broader view of modeling the estadístics. It aims to provide tools to automate (if possible) the whole process of data analysis and the “art” of statisticians selection hypothesis.
In conclusion, the KDD is an interdisciplinary field that provides an objectified knowledge to the professional and to the healthcare manager specifically in a more flexible and economical way, and provide a great support to the healthcare management for; decisions making and planning.
Remember that we must not confuse KDD with the data mining that is only a core passage of the KDD overall process.