Machine Learning Básico

#!/bin/bash # statistical_analysis.sh analyze_data() { local data="$1" awk ' BEGIN { print "=== Análise Estatística ===" } { sum += $1 sumsq += $1 * $1 values[NR] = $1 } END { # Média mean = sum/NR # Variância variance = sumsq/NR - mean^2 # Desvio Padrão stddev = sqrt(variance) # Mediana asort(values) if (NR % 2) median = values[int(NR/2) + 1] else median = (values[NR/2] + values[NR/2 + 1])/2 print "Amostras:", NR print "Média:", mean print "Mediana:", median print "Desvio Padrão:", stddev }' <<< "$data" }

#!/bin/bash # data_normalization.sh normalize_data() { local data="$1" local method="$2" # min-max ou z-score case "$method" in "min-max") awk ' NR == 1 { min = max = $1 } { if ($1 < min) min = $1 if ($1 > max) max = $1 values[NR] = $1 } END { for (i=1; i<=NR; i++) print (values[i] - min)/(max - min) }' <<< "$data" ;; "z-score") awk ' { sum += $1 sumsq += $1 * $1 values[NR] = $1 } END { mean = sum/NR stddev = sqrt(sumsq/NR - mean^2) for (i=1; i<=NR; i++) print (values[i] - mean)/stddev }' <<< "$data" ;; esac }

#!/bin/bash # kmeans.sh kmeans_clustering() { local data="$1" local k="$2" local max_iter=100 awk -v k="$k" -v max_iter="$max_iter" ' function dist(x1, y1, x2, y2) { return sqrt((x1-x2)^2 + (y1-y2)^2) } BEGIN { srand() } # Carrega dados { x[NR] = $1 y[NR] = $2 } END { # Inicializa centroides aleatoriamente for (i=1; i<=k; i++) { c_x[i] = x[int(rand() * NR)] c_y[i] = y[int(rand() * NR)] } # Iterações for (iter=1; iter<=max_iter; iter++) { # Associa pontos aos clusters changed = 0 for (i=1; i<=NR; i++) { min_dist = dist(x[i], y[i], c_x[1], c_y[1]) cluster[i] = 1 for (j=2; j<=k; j++) { d = dist(x[i], y[i], c_x[j], c_y[j]) if (d < min_dist) { min_dist = d cluster[i] = j } } } # Atualiza centroides for (i=1; i<=k; i++) { new_x = new_y = count = 0 for (j=1; j<=NR; j++) { if (cluster[j] == i) { new_x += x[j] new_y += y[j] count++ } } if (count > 0) { new_x /= count new_y /= count if (new_x != c_x[i] || new_y != c_y[i]) changed = 1 c_x[i] = new_x c_y[i] = new_y } } if (!changed) break } # Imprime resultados print "Clusters encontrados:" for (i=1; i<=NR; i++) print x[i], y[i], cluster[i] }' <<< "$data" }

#!/bin/bash # linear_regression.sh linear_regression() { local data="$1" awk ' { x[NR] = $1 y[NR] = $2 sum_x += $1 sum_y += $2 sum_xy += $1 * $2 sum_xx += $1 * $1 } END { # Calcula coeficientes n = NR slope = (n * sum_xy - sum_x * sum_y)/(n * sum_xx - sum_x * sum_x) intercept = (sum_y - slope * sum_x)/n # Calcula R² mean_y = sum_y/n ss_tot = ss_res = 0 for (i=1; i<=n; i++) { pred = slope * x[i] + intercept ss_tot += (y[i] - mean_y)^2 ss_res += (y[i] - pred)^2 } r2 = 1 - ss_res/ss_tot print "Coeficiente Angular:", slope print "Intercepto:", intercept print "R²:", r2 }' <<< "$data" }

#!/bin/bash # model_evaluation.sh evaluate_model() { local actual="$1" local predicted="$2" paste <(echo "$actual") <(echo "$predicted") | \ awk ' { error = $1 - $2 abs_error += abs(error) sq_error += error^2 if ($1 == $2) correct++ } END { print "=== Métricas de Avaliação ===" print "MAE:", abs_error/NR print "MSE:", sq_error/NR print "RMSE:", sqrt(sq_error/NR) print "Acurácia:", correct/NR * 100 "%" }' }

#!/bin/bash # cross_validation.sh cross_validate() { local data="$1" local k=5 # k-fold awk -v k="$k" ' { lines[NR] = $0 } END { # Embaralha dados for (i=NR; i>1; i--) { j = int(rand() * i) + 1 temp = lines[i] lines[i] = lines[j] lines[j] = temp } # K-fold CV fold_size = int(NR/k) for (fold=1; fold<=k; fold++) { start = (fold-1) * fold_size + 1 end = fold * fold_size print "=== Fold", fold, "===" print "Test indices:", start, "-", end } }' <<< "$data" }

# Implemente um classificador: # 1. Prepare os dados # 2. Treine o modelo # 3. Avalie performance # 4. Faça previsões

# Desenvolva análise preditiva: # 1. Analise correlações # 2. Construa modelo # 3. Valide resultados # 4. Otimize parâmetros