Errata (First Printing)

Chapter 4

Table 4.7: "0.029/1.362" should be "gain ratio: 0.029/1.557" and "0.021" should be "0.019"
(reported by Chiung-Wei Huang, Department of Electronic Engineering, Ching-Yun Institute of Technology, December 2002)

Line 34: "divide-and-conquer" should be "separate-and-conquer"

Lines 5-6: ", because otherwise the resulting item set would contain more than four items. These are" should be ". For example, we do not consider (A C D) and (B C D) because (A B C D) can also be generated from (A B C) and (A B D) and if these two are not candidate three-item sets then (A B C D) cannot be a candidate four-item set. This leaves"
(reported by J.C. Bioch, Department of Computer Science, Erasmus University Rotterdam, March 2001)

Chapter 5

Page 123, Line 25; Page 123, Line 28; Page 125, Line 5: "95%" should be "80%"
(reported by Ray Littler, Department of Statistics, University of Waikato, January 2000)

Lines 25-26: "73.3%" should be "73.2%" and "76.8%" should be "76.7%"
(reported by Juan Carlos Borr�s, August 2000)

Lines 28-29: "70%" should be "69.1%" and "81%" should be "80.1%"
(reported by Juan Carlos Borr�s, August 2000)

Line 10: "z" should be "-z"

Line 6: "[0.733,0.768]" should be "[0.732,0.767]" and "[0.70,0.81]" should be "[0.691,0.801]"
(reported by Juan Carlos Borr�s, August 2000)

Lines 8-9: "[0.65,1.02]" should be "[0.549,0.881]" and ", particularly the upper one!" should be "!"
(reported by Juan Carlos Borr�s, August 2000)

Line 10: "Probabilities, of course, can never be greater than 1." should be deleted
(reported by Juan Carlos Borr�s, August 2000)

Line 36: "repeating it all:" should be "repeating it at all:"

Line 2: the Greek character mu (denoting the true mean) should be replaced by an overlined x (denoting the mean of the sample)

Line 8: "pt_A+qt_A" should be "pt_A+qt_B" and "pf_A+qf_A" should be "pf_A+qf_B"

Table 5.5: precision should be "TP / (TP + FP)" instead of "TP / (TP + FN)"
(reported by Stuart Inglis, Intelligenesis Corporation, December 1999)

Line 11: "plotted as two" should be "plotted as"

Table 5.6: the denominator should be replaced by its square root in the calculation of the correlation coefficient
(reported by Huan Liu, Arizona State University, September 2000)

Chapter 6

Line 34: "the error rate could not possibly" should be replaced by "it would be a bad mistake to let the error rate"

Lines 1-2: ", which frequently exceed the maximum possible value of 0.5, underscore the fact that" should be replaced by " should be taken with a grain of salt because"

Line 23: "0.17%" should be replaced by "0.017%"
(reported by Xin Xu, University of Waikato, June 2001)

Figure 6.11, Line 39: "subtreEerror" should be "subtreeError"

Line 3: ")" missing before ","

Line 4: "of values that the set of attributes can assume" should be "of attributes"

Line 15: The square root in the denominator should only enclose pi

Chapter 7

Line 1: "up to k*2^k" should be "up to 2^k"

Line 2: "the effect of adding every attribute to" should be deleted

Line 29: "minimum number of instances" should be "minimum number of instances in the majority class"

Line 22: "given impurity function" should be "given additive impurity function"

Chapter 8

Table 8.1, Line 13: "-I" should be "-i"
(reported by Leonard Trigg, Intelligenesis Corporation, January 2000)

Table 8.2: last line in first column should be "-S"

Line 37: "jaws.classifiers.j48.J48" should be "weka.classifiers.j48.J48"
(reported by Leonard Trigg, Intelligenesis Corporation, November 1999)

Line 26: "* }" should be "}"
(reported by Leonard Trigg, Intelligenesis Corporation, November 1999)

Line 21: "weka.classifiers" should be "weka.core"