def leave_one_out(x, y, is_continuous):
"""
使用留一法验证法对模型进行评估
:return: 模型留一法验证下的正确率
"""
cnt = 0
for i in range(len(x)):
train_x = np.delete(x, i, 0)
train_y = np.delete(y, i, 0)
test_x = x[i]
test_y = y[i]
naive_bayes = NaiveBayes(train_x, train_y, is_continuous)
naive_bayes.train()
test_result = naive_bayes.inference(test_x)
if test_result == test_y:
cnt += 1
return cnt / len(x)
[1, 1, 0, 0, 1, 0, 0.437, 0.211],
[1, 1, 1, 1, 1, 0, 0.666, 0.091],
[0, 2, 2, 0, 2, 1, 0.243, 0.267],
[2, 2, 2, 2, 2, 0, 0.245, 0.057],
[2, 0, 0, 2, 2, 1, 0.343, 0.099],
[0, 1, 0, 1, 0, 0, 0.639, 0.161],
[2, 1, 1, 1, 0, 0, 0.657, 0.198],
[1, 1, 0, 0, 1, 1, 0.360, 0.370],
[2, 0, 0, 2, 2, 0, 0.593, 0.042],
[0, 0, 1, 1, 1, 0, 0.719, 0.103],
], dtype=object)
y = np.array([1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0])
is_continuous = [False, False, False, False, False, False, True, True]
naive_bayes = NaiveBayes(x, y, is_continuous)
naive_bayes.train()
print(naive_bayes.inference([0, 0, 0, 0, 0, 0, 0.697, 0.460]))
"""
y = 0
[0.33333333 0.25 0.41666667]
[0.33333333 0.41666667 0.25 ]
[0.41666667 0.33333333 0.25 ]
[0.25 0.41666667 0.33333333]
[0.25 0.33333333 0.41666667]
[0.63636364 0.36363636]
---
y = 1
[0.36363636 0.45454545 0.18181818]
[0.54545455 0.36363636 0.09090909]
[0.63636364 0.27272727 0.09090909]
[0.72727273 0.18181818 0.09090909]
def main():
raw_data = pd.read_csv("student_data.csv")
y = np.array(np.array(raw_data).T[0].T, dtype=int)
details = [
(1, 3, 4),
(2, 1, 3),
(3, 1, 4),
]
for feat, st, ed in details:
x = np.array(raw_data.iloc[:, st: ed])
is_continuous = [True] * feat
acc = leave_one_out(x, y, is_continuous)
print(f"accuracy using {feat} feature: {acc * 100}%")
fig = make_subplots(
rows=1, cols=2,
column_width=[0.6, 0.4],
specs=[
[{"type": "scatter3d"}, {"type": "scatter"}],
],
subplot_titles=(
"visualization of training data",
"ROC curve of models using different features",
)
)
fig.add_trace(
go.Scatter3d(
x=np.array(raw_data).T[1],
y=np.array(raw_data).T[2],
z=np.array(raw_data).T[3],
marker=dict(color=y), mode="markers",
showlegend=False,
),
row=1, col=1
)
fig.add_shape(
type="line", line=dict(dash="dash"),
x0=0, x1=1, y0=0, y1=1,
row=1, col=2,
)
for i in range(1, 4):
train_x = np.array(raw_data.iloc[:20, i: i + 1])
train_y = np.array(np.array(raw_data.iloc[:20, :]).T[0].T, dtype=int)
inference_x = np.array(raw_data.iloc[20:, i: i + 1])
inference_y = np.array(np.array(raw_data.iloc[20:, :]).T[0].T, dtype=int)
is_continuous = [True]
naive_bayes = NaiveBayes(train_x, train_y, is_continuous)
naive_bayes.inference = MethodType(inference, naive_bayes)
naive_bayes.train()
ans = []
tot, cnt = [0, 0], [0, 0]
for x_i, y_i in zip(inference_x, inference_y):
ans.append([naive_bayes.inference(x_i), y_i])
tot[y_i] += 1
ans.sort()
auc = 0.0
plot_x, plot_y = [0], [0]
for elem in ans:
cnt[elem[1]] += 1
plot_x.append(cnt[0] / tot[0])
plot_y.append(cnt[1] / tot[1])
if plot_x[-1] != plot_x[-2]:
auc += (plot_x[-1] - plot_x[-2]) * plot_y[-1]
fig.add_trace(
go.Scatter(
x=plot_x, y=plot_y,
name=f"feature {i} (AUC={round(auc, 4)})",
mode="lines",
),
row=1, col=2,
)
fig.update_xaxes(title_text="False Positive Rate", range=[-0.01, 1.01], row=1, col=2)
fig.update_yaxes(title_text="True Positive Rate", range=[-0.01, 1.01], row=1, col=2)
fig["layout"]["scene"]["xaxis"] = {"title": {"text": "height"}}
fig["layout"]["scene"]["yaxis"] = {"title": {"text": "weight"}}
fig["layout"]["scene"]["zaxis"] = {"title": {"text": "shoe size"}}
fig.show()
