The Students-Instructor Forum

Domanda classificator

Domanda classificator

by Alessandro Fossati -
Number of replies: 1

Buongiorno professore, ci chiedevamo quale fosse il corretto ragionamento da effettuare per rispondere a questa tipologia di domande, in quanto abbiamo riscontrato alcune incongruenze nello svolgere i quiz:

- Assume you learnt a classifier to predict whether a customer is loyal or not. Assume you used a data set with 1000 records where 900 are associated with loyal customers. Furthermore, assume you applied 10-folds cross validation and estimated the accuracy of your classifier to be 0.95. What can be said about your classifier?

- Assume you learnt a classifier to predict whether a customer is loyal or not. Assume you used a data set with 10000 records where 9900 are associated with loyal customers. Furthermore, assume you applied 10-folds cross validation and estimated the accuracy of your classifier to be 0.99. What can be said about your classifier?

Risposte possibili per entrambe:

- It is a poor classifier and will be ineffective to correctly predict which new costumers are loyal and which are not

- It is an excellent classifier and will be effective to predict which new costumers are loyal and which are not

- I can not make any statement about its usefulness with such a limited information about the problem, data set and estimated accuracy value

- it is an excellent classifier but nothing can be told about its effectiveness to predict which new customers are loyal and which are not



Grazie per l'attenzione



In reply to Alessandro Fossati

Re: Domanda classificator

by Fabio Antonio Stella -
Ricontrollate per bene la teoria e i concetti di intervalli di confidenza e di classi sbilanciate.
Buono studio