: :   data quality     : :   data morphing     : :   data mining     : :   decision trees     : :   bioinformatics






The aim of Data Mining
With the increase of database volume and the generalisation of data access, analysis needs of business and trade users created the demand for the development of Data Mining tools.

More than classical data navigation products, business users now also need analysis tools that can perform guided and on-line analysis, following an interactive and user-friendly process, and giving easy to understand results.


Data warehouse development
With the development of data warehouses and storage solutions, the amount of data held by companies is becoming more and more important.

Much of that information may not seem interesting at first, are stored because it is now easy and inexpensive. This point explains why data mining has become a key issue in today's information systems: when data mining is needed, the raw data necessary to process the analysis will almost always be there, waiting to reveal all the hidden trends it holds.


ISoft and Data Mining
ISoft provides tools that are designed specifically for the business user. With ALICE d'ISoft, you will be assisted during these steps. This user friendly tool with its intuitive interface will guide you through all the analyses, performed as an on-line process.

Use Amadea to prepare your data and Discovery to evaluate the quality of your source data.




The need for Data Mining
The need for data mining has developed since business users wanted to reappropriate their data, to easily test their model and build new ones, guided by their strong knowledge of business. One major improvement of data mining over classical navigation tools is that, more than just testing an already supposed trend, it can discover completely new patterns, and fully validate a model. The model can be built automatically but each step of its construction can be controlled by the business user. This model then enables him to predict the behaviour of other data, and to evaluate their chances of behaving as predicted.

But this work on the data is efficient only if it follows a good methodology. Several steps should be followed, in order to reach significant results.
Starting from heterogeneous raw data, you should first set up your data warehouse, in order to have a clean relational database.

You can then start the alimentation of your data mining tool. At this stage, you should be able to provide a good data description.

One key step is a good definition of your problematic. However powerful it might be, a data mining tool needs you to isolate the question you want to answer. Even if this question can change during the analysis, a good starting point must be defined.

Then only you can proceed to the analysis steps. During the analysis, you will need to go back several times to a preparation step, where you can define new, more relevant variables, use aggregation formulas... During these steps you will draw on your knowledge, to isolate and keep the interesting models and remove others..

After your analysis, you will be able to report your conclusions, as an expert in your field.


Back to HOME Back to HOME