update pkgs and re-run codes

Codes/Julia/Part-2/Project.toml

@@ -1,6 +1,12 @@
 [deps]
 CSV = "336ed68f-0bac-5ca0-87d4-7b16caf5d00b"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
+LaTeXStrings = "b964fa9f-0449-5b57-a5c2-d3ea65f4040f"
 MLJ = "add582a8-e3ab-11e8-2d5e-e98b27df1bc7"
+MLJClusteringInterface = "d354fa79-ed1c-40d4-88ef-b8c7bd1568af"
+MLJDecisionTreeInterface = "c6f25543-311c-4c74-83dc-3ea6d1015661"
+MLJLIBSVMInterface = "61c7150f-6c77-4bb1-949c-13197eac2a52"
+MLJLinearModels = "6ee0df7b-362f-4a72-a706-9e79364fb692"
+NearestNeighborModels = "636a865e-7cf4-491e-846c-de09b730eb36"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 ProgressMeter = "92933f4c-e287-5a05-a399-4b506db050ca"

Codes/Julia/Part-2/decision-tree-classification.jl

@@ -1,11 +1,12 @@
 ##################################
 #= DECISION TREE CLASSIFICATION =#
 ##################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
 md"Import librairies"
-using CSV, DataFrames
+using CSV, DataFrames, Plots
 using MLJ
 
 md"Read dataset and assign it `df`"

Codes/Julia/Part-2/decision-tree-regression.jl

@@ -1,6 +1,7 @@
 ##############################
 #= DECISION TREE REGRESSION =#
 ##############################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 

Codes/Julia/Part-2/gradient-descent.jl

@@ -1,6 +1,7 @@
 ###########################################
 # IMPLEMENTING GRADIENT DESCENT ALGORITHM #
 ###########################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 md"**NORMAL EQUATION**"
@@ -25,7 +26,7 @@ for _ in 1:1_000 # RUN IT MULTIPLE TIMES TO CONVERGE
     for k in shuffle(1:5)
 		cost = ( y[k] - X[k, :]' * θ )^2
 		push!(J, cost);
-		θ += α * (y[k] - X[k, :]' * θ) * X[k, :]
+		global θ += α * (y[k] - X[k, :]' * θ) * X[k, :]
 		println("θ is $(θ)")
     end
 end
@@ -39,10 +40,10 @@ for _ in 1:1_000
 	ϵ = (y-X*θ)[:, 1]
 	cost = (2*n)\ ϵ'*ϵ
 	push!(J, cost)
-	θ += n\α * sum((y - X * θ) .* X, dims=1)'
+	global θ += n\α * sum((y - X * θ) .* X, dims=1)'
 	println("θ is $(θ)")
 end
 
 plot(J)
 
-md"TO DO: Add Regularization Term To Cost Function"
+md"TO DO: Add Regularization Term To Cost Function"

Codes/Julia/Part-2/kmeans.jl

@@ -1,6 +1,7 @@
 ############
 #= KMEANS =#
 ############
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
@@ -31,7 +32,7 @@ kmeans_ = KMeans(k=5)
 md"You may want to see [Clustering.jl](https://github.com/JuliaStats/Clustering.jl) and the unwrapped model type [`Clustering.KMeans`](@ref)."
 
 md"Train & regroup into clusters"
-kmeans = machine(kmeans_, table(X)) |> fit!
+kmeans = machine(kmeans_, X) |> fit!
 
 md"Clusters & centroids"
 centroids = fitted_params(kmeans).centers

Codes/Julia/Part-2/knn.jl

@@ -1,6 +1,7 @@
 #########################
 #= k-NEAREST NEIGHBORS =#
 #########################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
@@ -49,5 +50,11 @@ ŷ = predict_mode(pipe, Xtest)
 md"Confusion Matrix"
 confusion_matrix(ŷ, ytest)
 
+md"Evaluation Metrics"
+accuracy(ŷ, ytest)
+specificity(ŷ, ytest) # specificity, true negative rate: TN/(TN+FP)
+sensitivity(ŷ, ytest) # sensitivity, true positive rate: TP/(TP+FN)
+f1score(ŷ, ytest)
+
 md"We can estimate the performance of `pipe` through the `evaluate!` command."
-evaluate!(pipe, operation=predict_mode, measures=[accuracy, precision, recall, f1score])
+evaluate!(pipe, operation=predict)

Codes/Julia/Part-2/logistic-regression.jl

@@ -1,6 +1,7 @@
 #########################
 #= LOGISTIC REGRESSION =#
 #########################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
@@ -15,7 +16,7 @@ schema(df)
 coerce!(df, 
         :Age => Continuous,
         :EstimatedSalary => Continuous,
-        :Purchased => Multiclass)
+        :Purchased => OrderedFactor)
 schema(df)
 
 md"Unpack features & target"
@@ -55,9 +56,9 @@ confusion_matrix(ŷ, ytest)
 
 md"Evaluation Metrics"
 accuracy(ŷ, ytest)
-precision(ŷ, ytest)
-recall(ŷ, ytest)
+specificity(ŷ, ytest) # specificity, true negative rate: TN/(TN+FP)
+sensitivity(ŷ, ytest) # sensitivity, true positive rate: TP/(TP+FN)
 f1score(ŷ, ytest)
 
 md"We can estimate, more elegantly, the performance of `pipe` through the `evaluate!` command."
-evaluate!(pipe, operation=predict_mode, measures=[accuracy, precision, recall, f1score])
+evaluate!(pipe, operation=predict)

Codes/Julia/Part-2/ml-workflows.jl

@@ -1,6 +1,7 @@
 ###########################################
 #= COMMON DATA PREPROCESSING `WORKFLOWS` =#
 ###########################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 

Codes/Julia/Part-2/multivariable-regression.jl

@@ -1,6 +1,7 @@
 #####################################
 #= MULTIVARIABLE LINEAR REGRESSION =#
 #####################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 

Codes/Julia/Part-2/polynomial-regression.jl

@@ -1,6 +1,8 @@
 ##################################
 #= POLYNOMIAL LINEAR REGRESSION =#
 ##################################
+# `versioninfo()` -> 1.11.1
+
 using Markdown
 
 md"Import librairies"

Codes/Julia/Part-2/random-forest-classification.jl

@@ -1,11 +1,12 @@
 ##################################
 #= RANDOM FOREST CLASSIFICATION =#
 ##################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
 md"Import librairies"
-using CSV, DataFrames
+using CSV, DataFrames, Plots
 using MLJ
 
 md"Read dataset -> `df`"
@@ -33,4 +34,4 @@ rfc = machine(rfc_, x, y) |> fit!
 md"You may want to see [DecisionTree.jl](https://github.com/bensadeghi/DecisionTree.jl) and the unwrapped model type [`MLJDecisionTreeInterface.DecisionTree.RandomForestClassifier`](@ref)."
 
 md"Evaluate the model"
-evaluate!(rfc)
+evaluate!(rfc)

Codes/Julia/Part-2/random-forest-regression.jl

@@ -1,12 +1,12 @@
 ##############################
 #= DECISION TREE REGRESSION =#
 ##############################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
 md"Import librairies"
-using CSV, D
-ataFrames
+using CSV, DataFrames
 using MLJ
 
 md"Load data from CSV file"

Codes/Julia/Part-2/simple-regression-1.jl

@@ -3,6 +3,8 @@
 ##############################
 # SALARY vs. YEARSEXPERIENCE #
 
+# `versioninfo()` -> 1.11.1
+
 using Markdown
 
 md"Import librairies"

Codes/Julia/Part-2/simple-regression-2.jl

@@ -3,6 +3,8 @@
 #######################
 #  WEIGHT vs. HEIGHT  #
 
+# `versioninfo()` -> 1.11.1
+
 using Markdown
 
 md"Import librairies"
@@ -42,4 +44,4 @@ md"Metric"
 println("Error is $(sum( (yhat .- ytest).^2 ) ./ length(ytest) )")
 
 scatter(xtest, ytest, label=:none)
-scatter!(xtest, yhat, label=:none)
+scatter!(xtest, yhat, label=:none)

Codes/Julia/Part-2/svc.jl

@@ -1,6 +1,7 @@
 ###############################################
 #= SUPPORT VECTOR MACHINE FOR CLASSIFICATION =#
 ###############################################
+# `versioninfo()` -> 1.11.1
 
 using Markdown
 
@@ -17,7 +18,7 @@ schema(df)
 coerce!(df, 
         :Age => Continuous,
         :EstimatedSalary => Continuous,
-        :Purchased => Multiclass)
+        :Purchased => OrderedFactor)
 schema(df)
 
 md"Unpack features & target"
@@ -49,5 +50,11 @@ ŷ = predict(pipe, Xtest)
 md"Confusion Matrix"
 confusion_matrix(ŷ, ytest)
 
+md"Evaluation Metrics"
+accuracy(ŷ, ytest)
+specificity(ŷ, ytest) # specificity, true negative rate: TN/(TN+FP)
+sensitivity(ŷ, ytest) # sensitivity, true positive rate: TP/(TP+FN)
+f1score(ŷ, ytest)
+
 md"We can estimate the performance of `pipe` through the `evaluate!` command."
-evaluate!(pipe, operation=predict, measures=[accuracy, precision, recall, f1score])
+evaluate!(pipe, operation=predict)