def dbscan(self, min_samples, eps=None, window=None):
self.window = window
'''
self.tickers = window.snp_live_tickers
flat_data = self.average_over_time(window)
self.normalised = StandardScaler().fit_transform(flat_data)
'''
fundamental = self.window.get_fundamental()
def get_ticker(index_list):
result = []
for i in index_list:
ticker = i[0]
if ticker not in result:
result.append(ticker)
return result
fundamental_tickers = get_ticker(fundamental.index)
num_of_feature = 7
fundamental_reshaped = fundamental.values.reshape(-1, num_of_feature)
fundamental_normalised = StandardScaler().fit_transform(
fundamental_reshaped)
fundamental_normalised = pd.DataFrame(fundamental_normalised,
index=fundamental_tickers)
fundamental_normalised = fundamental_normalised.dropna()
self.tickers = fundamental_normalised.index
dbscan = DBSCAN(eps=eps,
min_samples=min_samples).fit(fundamental_normalised)
self.dbscan_labels = labels = dbscan.labels_
core_samples_mask = np.zeros_like(dbscan.labels_, dtype=bool)
core_samples_mask[dbscan.core_sample_indices_] = True
self.dbscan_core_indices = dbscan.core_sample_indices_
self.dbscan_core_length = len(dbscan.core_sample_indices_)
self.dbscan_core_mask = core_samples_mask
self.unique_labels = set(labels)
self.n_clusters = n_clusters = len(
set(labels)) - (1 if -1 in labels else 0)
self.n_noise = list(labels).count(-1)
self.noise = np.where(labels == -1)[0]
clusters = {}
for j in range(n_clusters):
pairs = []
for i in itertools.combinations(np.where(labels == j)[0], 2):
pair = (i[0], i[1])
if window is not None:
pair = (self.tickers[i[0]], self.tickers[i[1]])
pairs.append(pair)
clusters[j] = pairs
pair_count = 0
all_pairs = clusters.values()
for i in all_pairs:
pair_count += len(i)
print('total pairs: ', pair_count)
return clusters
def fill_knn(df, train_cols, label_cols):
# del_cols = []
# for col in train_cols:
# if (df[col] == df[col][0]).all():
# del_cols.append(col)
# for col in del_cols:
# train_cols.remove(col)
data = StandardScaler().fit_transform(df.loc[:, train_cols].values)
data = pd.DataFrame(data, columns=train_cols)
data.dropna(axis=1, how='any', inplace=True)
knn_train_cols = [col for col in train_cols if col in data.columns]
for col in label_cols:
data_train = data.loc[df[col].notnull().tolist(),
knn_train_cols].values
data_label = df.loc[df[col].notnull().tolist(), col].values
knn = neighbors.KNeighborsRegressor()
knn.fit(data_train, data_label)
df.loc[df[col].isnull(),
col] = knn.predict(data.loc[df[col].isnull().tolist(),
knn_train_cols].values)
#optimum bilese sayisi
pca = PCA().fit(df)
plt.plot(np.cumsum(pca.explained_variance_ratio_))
plt.xlabel("Bileşen Sayısını")
plt.ylabel("Kümülatif Varyans Oranı")
pca.explained_variance_ratio_
#final
pca = PCA(n_components=3)
pca_fit = pca.fit_transform(df)
#burası açıklama oranı oluyor bu sayede o bilggilerle ne kadar açıklayıcı olabildiğimizi görebiliyoruz
pca.explained_variance_ratio_
#örenek yapıyorum
df = pd.read_csv("diabetes.csv", sep=",")
df = df.dropna()
dms = pd.get_dummies(df[['Age', 'DiabetesPedigreeFunction', 'Insulin']])
y = df["Outcome"]
#okunmayan değerleri silmem lazım
df
X_ = df.drop(['Outcome', 'Age', 'DiabetesPedigreeFunction', 'Insulin'],
axis=1).astype('float64')
X = pd.concat([X_, dms[['DiabetesPedigreeFunction', 'Insulin']]], axis=1)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.25,
random_state=42)