def treenimine(m):
algus = datetime.now()
print(f"\n\nN-grammidega ennustamine {n['tuup']} {n['tahis']}")
print(f"{time.ctime(time.time())}")
print(f"ngram size: {n['ngram_size']}, df size: {n['df_size']}")
print("andmepunktid: {0[0]}, tunnused: {0[1]}".format(vektorid.shape))
results = []
names = []
for name, model in m:
algus2 = datetime.now()
kfold = StratifiedKFold(n_splits=10, random_state=1, shuffle=True)
cv_results = cross_val_score(
model,
preprocessing.scale(X_train, with_mean=False),
np.ravel(Y_train),
cv=kfold,
scoring="accuracy",
n_jobs=-1,
)
results.append(cv_results)
names.append(name)
lõpp2 = datetime.now()
aeg2 = lõpp2 - algus2
print(
"{:40s} {:3.06f} {:10s} {:20s}".format(
f"{name}:",
round(cv_results.mean(), 6),
"({:1.06f})".format(round(cv_results.std(), 6)),
f"(Aeg: {str(aeg2)})",
)
)
lõpp = datetime.now()
aeg = lõpp - algus
print(f"Aeg: {aeg}\n\n")
pyplot.boxplot(results, labels=names)
pyplot.title("Algorithm Comparison / ngramm {} {}".format(n["tuup"], n["tahis"])),
pyplot.xticks(rotation=90)
pyplot.draw()
pyplot.savefig(
nimi(
"graafikud/{} eksimismaatriks ngramm {} {}".format(
name, n["tuup"], n["tahis"]
),
"png",
),
bbox_inches="tight",
dpi=100,
)
def funktsioon(ds, m):
algus = datetime.now()
array = ds.values
x = np.array(array[:, 1:])
y = np.array(array[:, :1])
scaler = preprocessing.StandardScaler().fit(x)
X_scaled = scaler.transform(x)
X_train, X_validation, Y_train, Y_validation = train_test_split(
X_scaled, y, test_size=0.20, shuffle=True
)
print("\n\nsõnatüübid")
print(f"{time.ctime(time.time())}")
print("andmepunktid: {}, tunnused: 17\n".format(len(X_train)))
for name, model in m:
algus2 = datetime.now()
model.fit(X_train, Y_train)
names = ["eesti", "soome", "vene"]
disp = plot_confusion_matrix(
model,
X_validation,
Y_validation,
display_labels=names,
cmap=pyplot.cm.Blues,
normalize="true",
)
disp.ax_.set_title(f"Normaliseeritud eksimismaatriks {name}")
pyplot.draw()
pyplot.savefig(
nimi("graafikud/Normaliseeritud eksimismaatriks {name} ", "png"),
bbox_inches="tight",
dpi=100,
)
lõpp2 = datetime.now()
aeg2 = lõpp2 - algus2
print(
"{:40s} {:150s} {:20s}".format(
f"{name}:",
f"{list(disp.confusion_matrix)}",
f"(Aeg: {str(aeg2)})",
)
)
lõpp = datetime.now()
aeg = lõpp - algus
print(f"Aeg: {aeg}\n\n")
y = np.ceil((df_suurused[1] - df_suurused[0]) / df_aste)
pyplot.xticks(
np.arange(x),
list(range(ngrammi_suurused[0], ngrammi_suurused[1], ngrammi_aste)),
)
pyplot.yticks(
np.arange(y), list(range(df_suurused[0], df_suurused[1], df_aste))
)
pyplot.title(
str(f"N-grammi tüüp: {ngrammi_tahis}, N-grammi tüüp {ngrammi_tuup}")
)
pyplot.draw()
pyplot.savefig(
nimi(
f"graafikud/täpsused ngramm {ngrammi_tahis} {ngrammi_tuup}", "png"
),
bbox_inches="tight",
dpi=100,
)
except TypeError:
print(
ngrammi_tahis,
ngrammi_tuup,
"#######################################",
toodeldud_andmed,
)
lõpp = datetime.now()
aeg = lõpp - algus
uus_tulemused = [[] for _ in range(len(models))]
for i in range(len(tulemused)):
for j in range(len(tulemused[i])):
uus_tulemused[j].append(tulemused[i][j])
laius = 3
korgus = 2
for j in range(len(uus_tulemused)):
fig, axs = pyplot.subplots(2, laius, figsize=(15, 10))
fig.suptitle(f"Normaliseeritud eksimismaatriks {nimed[j]}")
for i in range(len(uus_tulemused[0])):
axs[i // laius, i % laius].set_title(variandid[i])
df_cm = pd.DataFrame(uus_tulemused[j][i], index=keeled, columns=keeled)
sn.heatmap(df_cm,
annot=True,
cmap=pyplot.cm.Blues,
ax=axs[i // laius, i % laius])
pyplot.draw()
pyplot.savefig(
nimi(
f"graafikud/Normaliseeritud eksimismaatriks {nimed[j]} ngrammid",
"png",
),
bbox_inches="tight",
dpi=100,
)
pyplot.show()
pyplot.close()
def treenimine(ds, m):
algus = datetime.now()
array = ds.values
x = np.array(array[:, 1:])
y = np.array(array[:, :1])
scaler = preprocessing.StandardScaler().fit(x)
X_scaled = scaler.transform(x)
X_train, X_validation, Y_train, Y_validation = train_test_split(
X_scaled, y, test_size=0.20, shuffle=True)
results = []
names = []
print(f"\n\nSõnatüüpidel ennustamine")
print(f"{time.ctime(time.time())}")
print("andmepunktid: {}, tunnused: 17\n".format(len(X_train)))
for name, model in m:
algus2 = datetime.now()
kfold = StratifiedKFold(n_splits=10, shuffle=True)
try:
cv_results = cross_val_score(
model,
preprocessing.scale(X_train),
np.ravel(Y_train),
cv=kfold,
scoring="accuracy",
n_jobs=-1,
)
except:
print(
"================================| Error |================================"
)
results.append(cv_results)
names.append(name)
lõpp2 = datetime.now()
aeg2 = lõpp2 - algus2
print("{:40s} {:3.06f} {:10s} {:20s}".format(
f"{name}:",
round(cv_results.mean(), 6),
"({:1.06f})".format(round(cv_results.std(), 6)),
f"(Aeg: {str(aeg2)})",
))
pyplot.figure(figsize=(8, 6))
pyplot.boxplot(results, labels=names)
pyplot.title("Algoritmide täpsused")
pyplot.xticks(rotation=90)
pyplot.savefig(
nimi(
r"C:\Users\rasmu\OneDrive\Töölaud\Programmid\Python 3\Uurimistöö\Graafikud\Lõplikud\Sõnatüüpide efektiivsus ",
"png",
),
bbox_inches="tight",
dpi=100,
)
pyplot.draw()
lõpp = datetime.now()
aeg = lõpp - algus
print(f"Aeg: {aeg}\n\n")