non-numerical data

INTRODUCTION
almost any type of data can be included within this category
- social data
[ones from polls, surveys, human resources databases, etc.]
- categorised data (inherent scales)
[sensorial-like (hot, loud, etc.), colours, opinion-based scales (better, preferable, etc.)]
- categorisable data (can be fitted within arbitrary scales)
[locations, races, sex, religion, etc.]

TRENDINGBOT
before using the program with this data, it is required to perform a transformation into numerical values
rules
* low enough values
[the typical 0-10 scale works fine for the most of the cases]
* logical enough
[for eyes colours 0=brown & 10=green does not make sense brown-green-blue follows the logic evolution]
* max. & min. given by the raw data (don't extrapolate)
[for eyes colours and only brown & green in the raw data, the predictions wouldn´t be applicable to blue]

just an accessory tool

trendinBot intends to be a tool helping the human analyst (financial advisor, pollist, simulation engineer, etc.) to manage apparently unrelated information

hence, and excepting for the case of very simple situations, the actual version of the program should not be used as a standalone predictor

raw data & trendingbot

trendingBot is a user-friendly tool, however some to-the-point indications can help any newcomer to maximise its capabilities.

steps to be given in order to apply trendingbot to any set of data [program point of view]
1.- find the repetitive character [cases]
* it can be based on time periods, locations, persons, equivalent objects or situations, etc.
* some kind of consistency (regularity) has to be observed
2.- choose the variables whose values you want to predict [output for each case]
3.- select the most influential variables (up to 10) over the previous one [inputs for each case]

not sure?
run the program as many times as necessary and stick to the ones showing the highest accuracy

4.- run trendingBot and get the solution

dealing with data in a different way

(simplifying) statistical methods, although being based, in general, on sensible assumptions, imply an uncertainty, accepted until the moment as the less worse solution, but still is this the case?
nowadays, any home computer can easily perform calculations, difficult to be imagined just 10 ago. then? why simplifying further?

trendingBot offers an alternative path based on the following ideas:
1. simplifications = possible errors
2. the most complex situation can be divided into simpler behaviours
3. any of these behaviours can be mathematically described by selecting the appropriate variables

- error in the predictions means wrong variable selection
- any arbitrary user intervention (user-defined parameters) means wrong model delimitation

4. the most of the "natural" behaviours are based on quite simple mathematical relations
5. more detailed means better => combinatorics better than simplifying statical methods