IBM Switch 15 Uživatelský manuál stáhnout pdf (Strana 110)

100

Chapter 6

Figure 6-1

Specifying missing values for a continuous variable

Reading in mixed data.

Note that when y ou are reading in ﬁelds with numeric storage (either

integer, real, time, timest amp, or date), any non-numeric values are set to null or system missing.

This is because, unlike some applications, does not allow mixed storage types within a ﬁeld. To

avoid this, any ﬁelds with mix ed data should be read in as strings by changing the storage type in

the source node or exter nal application as necessa r y.

Reading empty strings from Oracle.

When reading from or writing to an Oracle database, be aware

that, unlike SPSS Modeler and unlike most othe r databases, Or acle treats and store s empty string

values as equivalent to null values. This means that the s ame data extracted from an Oracle

database may behave differently than when extracted from a ﬁle or another database, and the data

may return different results .

Handling Missing Values

You should decide how to treat missing values in li ght of your business or domain knowledge. To

ease training time and increase accuracy, you may w an t to remove blanks from your d ata set. On

the ot

her hand, the pres ence of blan k values may lea d to new business opp ortuniti es or a dditional

insights. In choosing the best technique, you should consider the following aspects of your data:

 Size of the data set

 Number of ﬁelds containing blanks

 Amount of missing inf ormation

1 2 ... 105 106 107 108 109 110 111 112 113 114 115 ... 269 270

Žádné komentáře

IBM Switch 15 Uživatelský manuál Strana 110