決定木モデルから各葉の規則を抽出する方法は？

Question

私は以下のようにpmml形式の意思決定ツリーモデルを持っています。各リーフのルールをテキストまたはその他の形式で保存するにはどうすればよいですか？例えば

：< = 3.5^clumpthickness < =^normalnucleoli> = 3.5 => B

<TreeModel modelName="DecisionTree" functionName="classification" splitCharacteristic="binarySplit" missingValueStrategy="lastPrediction" noTrueChildStrategy="returnNullPrediction"> 
    <MiningSchema> 
     <MiningField name="clumpthickness" invalidValueTreatment="asIs"/> 
     <MiningField name="uniformitycellsize" invalidValueTreatment="asIs"/> 
     <MiningField name="uniformitycellshape" invalidValueTreatment="asIs"/> 
     <MiningField name="marginaladhesion" invalidValueTreatment="asIs"/> 
     <MiningField name="epithelialcellsize" invalidValueTreatment="asIs"/> 
     <MiningField name="barenuclei" invalidValueTreatment="asIs"/> 
     <MiningField name="blandchromatin" invalidValueTreatment="asIs"/> 
     <MiningField name="normalnucleoli" invalidValueTreatment="asIs"/> 
     <MiningField name="mitoses" invalidValueTreatment="asIs"/> 
     <MiningField name="partition" invalidValueTreatment="asIs"/> 
     <MiningField name="Class_Categorical" invalidValueTreatment="asIs" usageType="target"/> 
    </MiningSchema> 
    <Node id="0" score="B" recordCount="559.0"> 
     <True/> 
     <ScoreDistribution value="B" recordCount="365.0"/> 
     <ScoreDistribution value="M" recordCount="194.0"/> 
     <Node id="1" score="B" recordCount="384.0"> 
     <SimplePredicate field="uniformitycellsize" operator="lessOrEqual" value="3.5"/> 
     <ScoreDistribution value="B" recordCount="356.0"/> 
     <ScoreDistribution value="M" recordCount="28.0"/> 
     <Node id="2" score="B" recordCount="368.0"> 
      <SimplePredicate field="clumpthickness" operator="lessOrEqual" value="6.5"/> 
      <ScoreDistribution value="B" recordCount="354.0"/> 
      <ScoreDistribution value="M" recordCount="14.0"/> 
      <Node id="3" score="B" recordCount="353.0"> 
      <SimplePredicate field="normalnucleoli" operator="lessOrEqual" value="3.5"/> 
      <ScoreDistribution value="B" recordCount="347.0"/> 
      <ScoreDistribution value="M" recordCount="6.0"/> 
      </Node> 
      <Node id="10" score="M" recordCount="15.0"> 
      <SimplePredicate field="normalnucleoli" operator="greaterThan" value="3.5"/> 
      <ScoreDistribution value="B" recordCount="7.0"/> 
      <ScoreDistribution value="M" recordCount="8.0"/> 
      </Node> 
     </Node> 
     <Node id="11" score="M" recordCount="16.0"> 
      <SimplePredicate field="clumpthickness" operator="greaterThan" value="6.5"/> 
      <ScoreDistribution value="B" recordCount="2.0"/> 
      <ScoreDistribution value="M" recordCount="14.0"/> 
     </Node> 
     </Node> 
     <Node id="12" score="M" recordCount="175.0"> 
     <SimplePredicate field="uniformitycellsize" operator="greaterThan" value="3.5"/> 
     <ScoreDistribution value="B" recordCount="9.0"/> 
     <ScoreDistribution value="M" recordCount="166.0"/> 
     <Node id="13" score="M" recordCount="33.0"> 
      <SimplePredicate field="uniformitycellsize" operator="lessOrEqual" value="4.5"/> 
      <ScoreDistribution value="B" recordCount="7.0"/> 
      <ScoreDistribution value="M" recordCount="26.0"/> 
      <Node id="14" score="M" recordCount="21.0"> 
      <SimplePredicate field="marginaladhesion" operator="lessOrEqual" value="5.5"/> 
      <ScoreDistribution value="B" recordCount="7.0"/> 
      <ScoreDistribution value="M" recordCount="14.0"/> 
      <Node id="15" score="B" recordCount="10.0"> 
       <SimplePredicate field="clumpthickness" operator="lessOrEqual" value="7.5"/> 
       <ScoreDistribution value="B" recordCount="6.0"/> 
       <ScoreDistribution value="M" recordCount="4.0"/> 
      </Node> 
      <Node id="16" score="M" recordCount="11.0"> 
       <SimplePredicate field="clumpthickness" operator="greaterThan" value="7.5"/> 
       <ScoreDistribution value="B" recordCount="1.0"/> 
       <ScoreDistribution value="M" recordCount="10.0"/> 
      </Node> 
      </Node> 
      <Node id="17" score="M" recordCount="12.0"> 
      <SimplePredicate field="marginaladhesion" operator="greaterThan" value="5.5"/> 
      <ScoreDistribution value="B" recordCount="0.0"/> 
      <ScoreDistribution value="M" recordCount="12.0"/> 
      </Node> 
     </Node> 
     <Node id="18" score="M" recordCount="142.0"> 
      <SimplePredicate field="uniformitycellsize" operator="greaterThan" value="4.5"/> 
      <ScoreDistribution value="B" recordCount="2.0"/> 
      <ScoreDistribution value="M" recordCount="140.0"/> 
     </Node> 
     </Node> 
    </Node> 
    </TreeModel> 

6.5 uniformitycellsize =================== ========================================== ====== このような結果を得るためのxslスタイルシートを以下に示します。

<xsl:stylesheet version="1.0" 
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> 
<xsl:output method="text" encoding="UTF-8"/> 

<xsl:template match="/"> 
    <xsl:for-each select="//Node[not(Node)]"> 
     <xsl:for-each select="ancestor-or-self::Node/SimplePredicate"> 
      <xsl:value-of select="@field"/> 
      <xsl:choose> 
       <xsl:when test="@operator = 'lessOrEqual'"> &lt;= </xsl:when> 
       <xsl:when test="@operator = 'greaterThan'"> &gt; </xsl:when> 
      </xsl:choose> 
      <xsl:value-of select="@value"/> 
      <xsl:if test="position() != last()"> 
       <xsl:text>^</xsl:text> 
      </xsl:if> 
      <xsl:if test="position() = last()"> 
       <xsl:text> => </xsl:text> 
       <xsl:value-of select="../@score"/> 
      </xsl:if> 
     </xsl:for-each> 
     <xsl:text>&#10;</xsl:text> 
    </xsl:for-each> 
</xsl:template> 

</xsl:stylesheet> 

出力結果は次のとおりです。

Uniformity of Cell Size <= 2.5^Bare Nuclei <= 5.5 => B 
Uniformity of Cell Size <= 2.5^Bare Nuclei > 5.5 => M 
Uniformity of Cell Size > 2.5^Uniformity of Cell Shape <= 2.5^Clump Thickness <= 5.5 => B 
Uniformity of Cell Size > 2.5^Uniformity of Cell Shape <= 2.5^Clump Thickness > 5.5 => M 
Uniformity of Cell Size > 2.5^Uniformity of Cell Shape > 2.5 => M 

Answer 1

は、XMLから葉を取得し、取得した情報からオブジェクトを構築するXPathを書くことができます。例えばnormalnucleoliため

XPathは次のようになります。上記のXPathを使用するには、Javaで//*[@field][@field='normalnucleoli']/@value

例のコードは次のようになります。あなたは、XMLを変換するために、JAXBを使用してアンマーシャラー書くことができます

DocumentBuilderFactory dbf = DocumentBuilderFactory.newInstance(); 
DocumentBuilder db = dbf.newDocumentBuilder(); 
Document doc = db.parse(new File(TreeModelXmlFile)); 
XPathFactory xPathFactory = XPathFactory.newInstance(); 
XPath xpath = xPathFactory.newXPath(); 


String fieldToExtract = "normalnucleoli"; 
String normalNucleoliValue = ""; 
XPathExpression expr = xpath.compile("//*[@field][@field='" + fieldToExtract + "']/@value"); 
Object exprEval = expr.evaluate(doc, XPathConstants.NODESET); 
if (exprEval != null && exprEval instanceof NodeList) 
{ 
    NodeList nodeList = (NodeList)exprEval; 
    if (nodeList.getLength() > 0) 
    { 
     normalNucleoliValue = nodeList.get(0).getTextContent(); 
    } 
}   
String operator = ""; 
expr = xpath.compile(""//*[@field][@field='" + fieldToExtract + "']/@operator""); 
Object exprEval = expr.evaluate(doc, XPathConstants.NODESET); 
if (exprEval != null && exprEval instanceof NodeList) 
{ 
    NodeList nodeList = (NodeList)exprEval; 
    if (nodeList.getLength() > 0) 
    { 
     operator = nodeList.get(0).getTextContent(); 
    }    
} 

System.out.println(fieldToExtract + " " + operator + " " + normalNucleoliValue); 

OR

javaオブジェクトに。それにはスキーマが必要です。あなたが何か行うことができXSLTで

Answer 2

：あなたの入力例に

XSLT 1.0

<xsl:stylesheet version="1.0" 
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> 
<xsl:output method="text" encoding="UTF-8"/> 

<xsl:template match="/"> 
    <xsl:for-each select="//Node[not(Node)]/ScoreDistribution"> 
     <xsl:for-each select="ancestor::Node/SimplePredicate"> 
      <xsl:value-of select="@field"/> 
      <xsl:choose> 
       <xsl:when test="@operator = 'lessOrEqual'"> &lt;= </xsl:when> 
       <xsl:when test="@operator = 'greaterThan'"> &gt; </xsl:when> 
      </xsl:choose> 
      <xsl:value-of select="@value"/> 
      <xsl:if test="position() != last()"> 
       <xsl:text>^</xsl:text> 
      </xsl:if> 
     </xsl:for-each> 
     <xsl:text> => </xsl:text> 
     <xsl:value-of select="@value"/> 
     <xsl:text> (</xsl:text> 
     <xsl:value-of select="@recordCount"/> 
     <xsl:text>) &#10;</xsl:text> 
    </xsl:for-each> 
</xsl:template> 

</xsl:stylesheet> 

応用し、その結果は以下のようになります。

uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli <= 3.5 => B (347.0) 
uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli <= 3.5 => M (6.0) 
uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli > 3.5 => B (7.0) 
uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli > 3.5 => M (8.0) 
uniformitycellsize <= 3.5^clumpthickness > 6.5 => B (2.0) 
uniformitycellsize <= 3.5^clumpthickness > 6.5 => M (14.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness <= 7.5 => B (6.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness <= 7.5 => M (4.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness > 7.5 => B (1.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness > 7.5 => M (10.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion > 5.5 => B (0.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion > 5.5 => M (12.0) 
uniformitycellsize > 3.5^uniformitycellsize > 4.5 => B (2.0) 
uniformitycellsize > 3.5^uniformitycellsize > 4.5 => M (140.0) 

---

あるいは、ご希望の場合：

<xsl:stylesheet version="1.0" 
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> 
<xsl:output method="text" encoding="UTF-8"/> 

<xsl:template match="/"> 
    <xsl:for-each select="//Node[not(Node)]"> 
     <xsl:for-each select="ancestor-or-self::Node/SimplePredicate"> 
      <xsl:value-of select="@field"/> 
      <xsl:choose> 
       <xsl:when test="@operator = 'lessOrEqual'"> &lt;= </xsl:when> 
       <xsl:when test="@operator = 'greaterThan'"> &gt; </xsl:when> 
      </xsl:choose> 
      <xsl:value-of select="@value"/> 
      <xsl:if test="position() != last()"> 
       <xsl:text>^</xsl:text> 
      </xsl:if> 
     </xsl:for-each> 
     <xsl:text> => </xsl:text> 
     <xsl:for-each select="ScoreDistribution"> 
      <xsl:value-of select="@value"/> 
      <xsl:text> (</xsl:text> 
      <xsl:value-of select="@recordCount"/> 
      <xsl:text>)</xsl:text> 
      <xsl:if test="position() != last()"> 
       <xsl:text>; </xsl:text> 
      </xsl:if> 
     </xsl:for-each> 
     <xsl:text>&#10;</xsl:text> 
    </xsl:for-each> 
</xsl:template> 

</xsl:stylesheet> 

生成する：

uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli <= 3.5 => B (347.0); M (6.0) 
uniformitycellsize <= 3.5^clumpthickness <= 6.5^normalnucleoli > 3.5 => B (7.0); M (8.0) 
uniformitycellsize <= 3.5^clumpthickness > 6.5 => B (2.0); M (14.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness <= 7.5 => B (6.0); M (4.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion <= 5.5^clumpthickness > 7.5 => B (1.0); M (10.0) 
uniformitycellsize > 3.5^uniformitycellsize <= 4.5^marginaladhesion > 5.5 => B (0.0); M (12.0) 
uniformitycellsize > 3.5^uniformitycellsize > 4.5 => B (2.0); M (140.0)