From cf60c03db2481d3816e63f058a7ed12d905ac833 Mon Sep 17 00:00:00 2001 From: mguetlein Date: Fri, 25 Nov 2011 09:07:50 +0100 Subject: add r-square plot, fix prediction updating, add weighted sample-correlation-coefficient --- lib/predictions.rb | 173 +++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 147 insertions(+), 26 deletions(-) (limited to 'lib') diff --git a/lib/predictions.rb b/lib/predictions.rb index 6c0e996..56bdd22 100755 --- a/lib/predictions.rb +++ b/lib/predictions.rb @@ -25,9 +25,6 @@ module Lib feature_type, accept_values=nil ) - @predicted_values = predicted_values - @actual_values = actual_values - @confidence_values = confidence_values @feature_type = feature_type @accept_values = accept_values @num_classes = 1 @@ -38,34 +35,27 @@ module Lib raise "unknown feature_type: '"+@feature_type.to_s+"'" unless @feature_type=="classification" || @feature_type=="regression" - raise "no predictions" if @predicted_values.size == 0 - num_info = "predicted:"+@predicted_values.size.to_s+ - " confidence:"+@confidence_values.size.to_s+" actual:"+@actual_values.size.to_s - raise "illegal num actual values "+num_info if @actual_values.size != @predicted_values.size - raise "illegal num confidence values "+num_info if @confidence_values.size != @predicted_values.size - - @confidence_values.each{ |c| raise "illegal confidence value: '"+c.to_s+"'" unless c==nil or (c.is_a?(Numeric) and c>=0 and c<=1) } + raise "no predictions" if predicted_values.size == 0 + num_info = "predicted:"+predicted_values.size.to_s+ + " confidence:"+confidence_values.size.to_s+" actual:"+actual_values.size.to_s + raise "illegal num actual values "+num_info if actual_values.size != predicted_values.size + raise "illegal num confidence values "+num_info if confidence_values.size != predicted_values.size case @feature_type when "classification" raise "accept_values missing while performing classification" unless @accept_values @num_classes = @accept_values.size raise "num classes < 2" if @num_classes<2 - { "predicted"=>@predicted_values, "actual"=>@actual_values }.each do |s,values| - values.each{ |v| raise "illegal "+s+" classification-value ("+v.to_s+"),"+ - "has to be either nil or index of predicted-values" if v!=nil and (!v.is_a?(Numeric) or v<0 or v>@num_classes)} - end when "regression" raise "accept_values != nil while performing regression" if @accept_values - { "predicted"=>@predicted_values, "actual"=>@actual_values }.each do |s,values| - values.each{ |v| raise "illegal "+s+" regression-value ("+v.to_s+"),"+ - " has to be either nil or number (not NaN, not Infinite)" unless v==nil or (v.is_a?(Numeric) and !v.nan? and v.finite?)} - end end + @predicted_values = [] + @actual_values = [] + @confidence_values = [] init_stats() - (0..@predicted_values.size-1).each do |i| - update_stats( @predicted_values[i], @actual_values[i], @confidence_values[i] ) + (0..predicted_values.size-1).each do |i| + update_stats( predicted_values[i], actual_values[i], confidence_values[i] ) end end @@ -114,6 +104,13 @@ module Lib @sum_squares_actual = 0 @sum_squares_predicted = 0 + @sum_confidence = 0 + @weighted_sum_actual = 0 + @weighted_sum_predicted = 0 + @weighted_sum_multiply = 0 + @weighted_sum_squares_actual = 0 + @weighted_sum_squares_predicted = 0 + @sum_weighted_abs_error = 0 @sum_weighted_squared_error = 0 end @@ -121,6 +118,25 @@ module Lib def update_stats( predicted_value, actual_value, confidence_value ) + raise "illegal confidence value: '"+confidence_value.to_s+"'" unless + confidence_value==nil or (confidence_value.is_a?(Numeric) and confidence_value>=0 and confidence_value<=1) + case @feature_type + when "classification" + { "predicted"=>predicted_value, "actual"=>actual_value }.each do |s,v| + raise "illegal "+s+" classification-value ("+v.to_s+"),"+ + "has to be either nil or index of predicted-values" if v!=nil and (!v.is_a?(Numeric) or v<0 or v>@num_classes) + end + when "regression" + { "predicted"=>predicted_value, "actual"=>actual_value }.each do |s,v| + raise "illegal "+s+" regression-value ("+v.to_s+"),"+ + " has to be either nil or number (not NaN, not Infinite)" unless v==nil or (v.is_a?(Numeric) and !v.nan? and v.finite?) + end + end + + @predicted_values << predicted_value + @actual_values << actual_value + @confidence_values << confidence_value + if actual_value==nil @num_no_actual_value += 1 else @@ -165,6 +181,16 @@ module Lib @sum_multiply += (actual_value*predicted_value) @sum_squares_actual += actual_value**2 @sum_squares_predicted += predicted_value**2 + + if @conf_provided + w_a = actual_value * confidence_value + w_p = predicted_value * confidence_value + @weighted_sum_actual += w_a + @weighted_sum_predicted += w_p + @weighted_sum_multiply += (w_a*w_p) if @conf_provided + @weighted_sum_squares_actual += w_a**2 if @conf_provided + @weighted_sum_squares_predicted += w_p**2 if @conf_provided + end end end end @@ -514,7 +540,7 @@ module Lib return @sum_squared_error end - def r_square + def r_square #_old #return sample_correlation_coefficient ** 2 # see http://en.wikipedia.org/wiki/Coefficient_of_determination#Definitions @@ -525,7 +551,7 @@ module Lib ( r_2.infinite? || r_2.nan? ) ? 0 : r_2 end - def weighted_r_square + def weighted_r_square #_old return 0 unless confidence_values_available? ss_tot = weighted_total_sum_of_squares return 0 if ss_tot==0 @@ -533,6 +559,16 @@ module Lib ( r_2.infinite? || r_2.nan? ) ? 0 : r_2 end + #def r_square + # # as implemted in R + # return sample_correlation_coefficient ** 2 + #end + + #def weighted_r_square + # # as implemted in R + # return weighted_sample_correlation_coefficient ** 2 + #end + def sample_correlation_coefficient begin # formula see http://en.wikipedia.org/wiki/Correlation_and_dependence#Pearson.27s_product-moment_coefficient @@ -543,6 +579,16 @@ module Lib rescue; 0; end end + def weighted_sample_correlation_coefficient + begin + # formula see http://en.wikipedia.org/wiki/Correlation_and_dependence#Pearson.27s_product-moment_coefficient + scc = ( @num_predicted * @weighted_sum_multiply - @weighted_sum_actual * @weighted_sum_predicted ) / + ( Math.sqrt( @num_predicted * @weighted_sum_squares_actual - @weighted_sum_actual**2 ) * + Math.sqrt( @num_predicted * @weighted_sum_squares_predicted - @weighted_sum_predicted**2 ) ) + ( scc.infinite? || scc.nan? ) ? 0 : scc + rescue; 0; end + end + def total_sum_of_squares #return @variance_actual * ( @num_predicted - 1 ) sum = 0 @@ -608,17 +654,23 @@ module Lib return h end - def get_prediction_values(actual_accept_value, predicted_accept_value) + def get_prediction_values(performance_attr, performance_accept_value) #puts "get_roc_values for class_value: "+class_value.to_s raise "no confidence values" unless confidence_values_available? #raise "no class-value specified" if class_value==nil + actual_accept_value = nil + predicted_accept_value = nil + if performance_attr==:true_positive_rate + actual_accept_value = performance_accept_value + elsif performance_attr==:positive_predictive_value + predicted_accept_value = performance_accept_value + end actual_class_index = @accept_values.index(actual_accept_value) if actual_accept_value!=nil raise "class not found '"+actual_accept_value.to_s+"' in "+@accept_values.inspect if (actual_accept_value!=nil && actual_class_index==nil) - predicted_class_index = @accept_values.index(predicted_accept_value) if predicted_accept_value!=nil - raise "class not found "+predicted_accept_value.to_s+" in "+@accept_values.inspect if (predicted_accept_value!=nil && predicted_class_index==nil) + raise "class not found '"+predicted_accept_value.to_s+"' in "+@accept_values.inspect if (predicted_accept_value!=nil && predicted_class_index==nil) c = []; p = []; a = [] (0..@predicted_values.size-1).each do |i| @@ -697,6 +749,67 @@ module Lib #end private + def self.test_update + p=[0.4,0.2,0.3,0.5,0.8] + a=[0.45,0.21,0.25,0.55,0.75] + c = Array.new(p.size) + pred = Predictions.new(p,a,c,"regression") + puts pred.r_square + + pred = nil + p.size.times do |i| + if pred==nil + pred = Predictions.new([p[0]],[a[0]],[c[0]],"regression") + else + pred.update_stats(p[i],a[i],c[i]) + end + puts pred.r_square + end + end + + def self.test_r_square + require "rubygems" + require "opentox-ruby" + + max_deviation = rand * 0.9 + avg_deviation = max_deviation * 0.5 + + p = [] + a = [] + c = [] + (100 + rand(1000)).times do |i| + r = rand + deviation = rand * max_deviation + a << r + p << r + ((rand<0.5 ? -1 : 1) * deviation) + #c << 0.5 + if (deviation > avg_deviation) + c << 0.4 + else + c << 0.6 + end + #puts a[-1].to_s+" "+p[-1].to_s + end + puts "num values "+p.size.to_s + + pred = Predictions.new(p,a,c,"regression") + puts "internal" + #puts "r-square old "+pred.r_square_old.to_s + puts "cor "+pred.sample_correlation_coefficient.to_s + puts "weighted cor "+pred.weighted_sample_correlation_coefficient.to_s + puts "r-square "+pred.r_square.to_s + + puts "R" + @@r = RinRuby.new(true,false) unless defined?(@@r) and @@r + @@r.assign "v1",a + @@r.assign "v2",p + puts "r cor "+@@r.pull("cor(v1,v2)").to_s + @@r.eval "fit <- lm(v1 ~ v2)" + @@r.eval "sum <- summary(fit)" + puts "r r-square "+@@r.pull("sum$r.squared").to_s + puts "r adjusted-r-square "+@@r.pull("sum$adj.r.squared").to_s + end + def prediction_feature_value_map(proc) res = {} (0..@num_classes-1).each do |i| @@ -706,4 +819,12 @@ module Lib end end -end \ No newline at end of file +end + +#class Float +# def to_s +# "%.5f" % self +# end +#end +##Lib::Predictions.test_update +#Lib::Predictions.test_r_square -- cgit v1.2.3