wink-regression-tree - npm Package Compare versions

Comparing version 1.1.0 to 1.1.1

package.json

 {
   "name": "wink-regression-tree",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "Decision Tree to predict the value of a continuous target variable",
@@ -35,3 +35,3 @@ "keywords": [
     "chai": "^4.1.0",
-    "coveralls": "^2.11.15",
+    "coveralls": "^3.0.0",
     "docker": "^1.0.0",
@@ -42,3 +42,3 @@ "documentation": "^5.3.1",
     "jshint": "^2.9.4",
-    "mocha": "^3.1.0",
+    "mocha": "^4.0.1",
     "mocha-lcov-reporter": "^1.2.0"
@@ -45,0 +45,0 @@ },

README.md

@@ -9,3 +9,3 @@ # wink-regression-tree
 
-Predict the value of a continuous variable such as price, turn around time, or mileage using **`wink-regression-tree`**. It is a part of _[wink](https://www.npmjs.com/~sanjaya)_ — a growing family of high quality packages for Statistical Analysis, Natural Language Processing and Machine Learning in NodeJS.
+Predict the value of a continuous variable such as price, turn around time, or mileage using **`wink-regression-tree`**. It is a part of _[wink](http://wink.org.in/)_ — a growing family of high quality packages for Statistical Analysis, Natural Language Processing and Machine Learning in NodeJS.
 
@@ -97,3 +97,3 @@
 ### Documentation
-For detailed API docs, check out https://winkjs.github.io/wink-regression-tree/ URL!
+For detailed API docs, check out http://wink.org.in/wink-regression-tree/ URL!
 
@@ -106,5 +106,5 @@ ### Need Help?
 
-**wink-regression-tree** is copyright 2017 GRAYPE Systems Private Limited.
+**wink-regression-tree** is copyright 2017 [GRAYPE Systems Private Limited](http://graype.in/).
 
 It is licensed under the under the terms of the GNU Affero General Public License as published by the Free
 Software Foundation, version 3 of the License.

src/wink-regression-tree.js

@@ -27,2 +27,3 @@ //     wink-regression-tree
 var helpers = require( 'wink-helpers' );
+var stdevEPSILON = Math.pow( 2, -48 );
 
@@ -380,4 +381,3 @@ // ### regressionTree
     if ( depth > config.maxDepth ) {
-      // Yes, Incrment rules learned & return.
-      wrTree.rulesLearned += 1;
+      // Yes, return.
       return;
@@ -397,3 +397,3 @@ }
       if ( splitData[ uniqVal ].size < config.minLeafNodeItems ) {
-        // Don't increment rules learned as you are pruning tree; just skip this iteration.
+        // Just skip this iteration.
         continue;
@@ -411,5 +411,4 @@ }
       // Does it have enough items to proceed with split?
-      if ( index.length <= config.minSplitCandidateItems ) {
+      if ( index.length <= config.minSplitCandidateItems || child.stdev < stdevEPSILON ) {
         // No! continue with the iteration with the next `uniqVal`.
-        wrTree.rulesLearned += 1;
         continue;
@@ -424,3 +423,2 @@ }
         // No best column found, coninue with the next one!
-        wrTree.rulesLearned += 1;
         continue;
@@ -432,3 +430,2 @@ }
         // No! continue with the iteration with the next `uniqVal`.
-        wrTree.rulesLearned += 1;
         continue;
@@ -449,2 +446,21 @@ }
 
+  // ### countRules
+  /**
+   *
+   * Counts the number of rules generated from a rules tree and updates the final
+   * number in the root node of the tree.
+   *
+   * @param {object} tree — the rules tree.
+   * @return {undefined} or void!
+   * @private
+  */
+  var countRules = function ( tree ) {
+    var subTree = tree.branches;
+    for ( var node in subTree ) {
+      if ( subTree[ node ].branches !== undefined && Object.keys( subTree[ node ].branches ).length > 0 ) {
+        countRules( subTree[ node ] );
+      } else wrTree.rulesLearned += 1;
+    }
+  }; // countRules()
+
   // ### defineConfig
@@ -608,17 +624,21 @@ /**
     wrTree.stdev = computeStdev( rootsVarianceXn, wrTree.size );
-    bestSplit = selectBestSplit( cndts );
-    if ( bestSplit === undefined ) {
-      // Opps, no worthy column available - return the root!
-      wrTree.rulesLearned += 1;
-      return true;
+    // Attempt to grow tree if standard deviation is large enough!
+    if ( wrTree.stdev > stdevEPSILON ) {
+      bestSplit = selectBestSplit( cndts );
+      if ( bestSplit === undefined ) {
+        // Opps, no worthy column available - return the root!
+        return true;
+      }
+      // Find the updated list of candidate columsn after the split.
+      for ( i = 0; i < candidateCols.length; i += 1 ) {
+        if ( candidateCols[ i ] !== bestSplit.col ) updatedCandidateCols.push( candidateCols[ i ] );
+      }
+      // Define the balance stuff as a split has been found!
+      wrTree.colUsed4Split = columnsDefn[xc2cMap[bestSplit.col]].name;
+      wrTree.varianceReduction = computePercentageVarianceReduction( rootsVarianceXn, wrTree.size, bestSplit.sum );
+      // Call recursive function, `growTree()`.
+      growTree( updatedCandidateCols, cndts.columns[ bestSplit.col ], bestSplit.col, wrTree, 1 );
     }
-    // Find the updated list of candidate columsn after the split.
-    for ( i = 0; i < candidateCols.length; i += 1 ) {
-      if ( candidateCols[ i ] !== bestSplit.col ) updatedCandidateCols.push( candidateCols[ i ] );
-    }
-    // Define the balance stuff as a split has been found!
-    wrTree.colUsed4Split = columnsDefn[xc2cMap[bestSplit.col]].name;
-    wrTree.varianceReduction = computePercentageVarianceReduction( rootsVarianceXn, wrTree.size, bestSplit.sum );
-    // Call recursive function, `growTree()`.
-    growTree( updatedCandidateCols, cndts.columns[ bestSplit.col ], bestSplit.col, wrTree, 1 );
+    wrTree.rulesLearned = 0;
+    countRules( wrTree );
     return wrTree.rulesLearned;
@@ -625,0 +645,0 @@ }; // learn()

No alert changes

Improved metrics

Total package byte prevSize

123326

0.49%

Lines of code

1174

1.65%

No dependency changes