[docs] document CLI behavior when label_column is omitted

[jameslamb]

Answering this stack overflow question tonight, I wasn't able to figure out from LightGBM's documentation that, when using the CLI, if label_column is omitted from train.conf, LightGBM defaults to using the first column in the training data as the target.

I had to figure this out from source code

• Dataset.label_idx_ is initialized to 0:

  LightGBM/src/io/dataset_loader.cpp

  Line 22 in a8ee487

  label_idx_ = 0;

• Dataset.label_idx_ is only overridden if label_column is provided in the config:

  LightGBM/src/io/dataset_loader.cpp

  Lines 48 to 57 in a8ee487

  	if (config_.label_column.size() > 0) {
  	if (Common::StartsWith(config_.label_column, name_prefix)) {
  	std::string name = config_.label_column.substr(name_prefix.size());
  	label_idx_ = -1;
  	for (int i = 0; i < static_cast<int>(feature_names_.size()); ++i) {
  	if (name == feature_names_[i]) {
  	label_idx_ = i;
  	break;
  	}
  	}

This PR proposes two small changes to make this clearer for new users of the CLI:

• adds a line to the docs for https://lightgbm.readthedocs.io/en/latest/Parameters.html#label_column explaining this behavior
• adds label_column to all of the training configs in examples/, to nudge users who copy those configs as a reference to explicitly define label_column in their own configs

[StrikerRUS]

Nice clarification!

[github-actions]

This pull request has been automatically locked since there has not been any recent activity since it was closed. To start a new related discussion, open a new issue at https://github.com/microsoft/LightGBM/issues including a reference to this.