add internal_value_ in tree and output to txt model, so we can use for...

add internal_value_ in tree and output to txt model, so we can use for calculating feature importance when predicting (#80)

include/LightGBM/tree.h

@@ -145,6 +145,8 @@ private:
   int* leaf_parent_;
   /*! \brief Output of leaves */
   double* leaf_value_;
+  /*! \brief Output of internal nodes(save internal output for per inference feature importance calc) */
+  double* internal_value_;
   /*! \brief Depth for leaves */
   int* leaf_depth_;
 };

src/io/tree.cpp

@@ -28,6 +28,7 @@ Tree::Tree(int max_leaves)
 
   leaf_parent_ = new int[max_leaves_];
   leaf_value_ = new double[max_leaves_];
+  internal_value_ = new double[max_leaves_ - 1];
   leaf_depth_ = new int[max_leaves_];
   // root is in the depth 1
   leaf_depth_[0] = 1;
@@ -44,6 +45,7 @@ Tree::~Tree() {
   if (threshold_ != nullptr) { delete[] threshold_; }
   if (split_gain_ != nullptr) { delete[] split_gain_; }
   if (leaf_value_ != nullptr) { delete[] leaf_value_; }
+  if (internal_value_ != nullptr) { delete[] internal_value_; }
   if (leaf_depth_ != nullptr) { delete[] leaf_depth_; }
 }
 
@@ -72,6 +74,8 @@ int Tree::Split(int leaf, int feature, unsigned int threshold_bin, int real_feat
   // update new leaves
   leaf_parent_[leaf] = new_node_idx;
   leaf_parent_[num_leaves_] = new_node_idx;
+  // save current leaf value to internal node before change
+  internal_value_[new_node_idx] = leaf_value_[leaf];
   leaf_value_[leaf] = left_value;
   leaf_value_[num_leaves_] = right_value;
   // update leaf depth
@@ -125,6 +129,8 @@ std::string Tree::ToString() {
     << Common::ArrayToString<int>(leaf_parent_, num_leaves_, ' ') << std::endl;
   ss << "leaf_value="
     << Common::ArrayToString<double>(leaf_value_, num_leaves_, ' ') << std::endl;
+  ss << "internal_value="
+    << Common::ArrayToString<double>(internal_value_, num_leaves_ - 1, ' ') << std::endl;
   ss << std::endl;
   return ss.str();
 }
@@ -145,7 +151,8 @@ Tree::Tree(const std::string& str) {
   if (key_vals.count("num_leaves") <= 0 || key_vals.count("split_feature") <= 0
     || key_vals.count("split_gain") <= 0 || key_vals.count("threshold") <= 0
     || key_vals.count("left_child") <= 0 || key_vals.count("right_child") <= 0
-    || key_vals.count("leaf_parent") <= 0 || key_vals.count("leaf_value") <= 0) {
+    || key_vals.count("leaf_parent") <= 0 || key_vals.count("leaf_value") <= 0
+    || key_vals.count("internal_value") <= 0) {
     Log::Fatal("Tree model string format error");
   }
 
@@ -158,6 +165,7 @@ Tree::Tree(const std::string& str) {
   split_gain_ = new double[num_leaves_ - 1];
   leaf_parent_ = new int[num_leaves_];
   leaf_value_ = new double[num_leaves_];
+  internal_value_ = new double[num_leaves_ - 1];
 
   split_feature_ = nullptr;
   threshold_in_bin_ = nullptr;
@@ -177,6 +185,8 @@ Tree::Tree(const std::string& str) {
                            num_leaves_ , leaf_parent_);
   Common::StringToDoubleArray(key_vals["leaf_value"], ' ',
                               num_leaves_ , leaf_value_);
+  Common::StringToDoubleArray(key_vals["internal_value"], ' ',
+                              num_leaves_ - 1 , internal_value_);
 }
 
 }  // namespace LightGBM