def pca_therapy_activities(therapy_filepath="data/therapyByUser.json"):
"""
Activity observations with PCA described by users' frequency
Color activities depending on their therapy
"""
therapy_activity = therapy_from_activity()
df = dataframe_activity_frequency()
df = df.transpose()
index = df.index
# print(df.describe())
pca = PCA(n_components=2)
pca.fit(df)
# print(pca.components_)
# print(pca.explained_variance_ratio_)
activity_2 = pca.transform(df)
activity_2 = pd.DataFrame(
{
"PCA1": activity_2[:, 0],
"PCA2": activity_2[:, 1],
"labels": [therapy_activity[activity] for activity in df.index]
},
index=index)
data = [
go.Scatter(x=activity_2[activity_2["labels"] == label]["PCA1"],
y=activity_2[activity_2["labels"] == label]["PCA2"],
mode="markers",
name=str(label),
marker={"size": 15},
text=activity_2[activity_2["labels"] == label].index)
for label in activity_2["labels"].unique()
]
py.plot(data)
def tsne_kmeans_users(n_clusters=3,
frequency_filepath="data/data_frequency.json"):
"""
Activity observations with TSNE described by users' frequency
Compute K-means algorithm
:param n_clusters int (optional): Number of clusters for K-means algorithm
"""
df_freq = dataframe_activity_frequency()
# print(df_freq.head())
tsne = TSNE(n_components=2)
users_2 = tsne.fit_transform(df_freq)
# print(pca.components_)
# print(pca.explained_variance_ratio_)
# data = [go.Scatter(
# x=users_2[:,0],
# y=users_2[:,1],
# mode = "markers"
# )]
kmeans = KMeans(n_clusters=2).fit(df_freq)
labels = kmeans.labels_
activity_2 = pd.DataFrame(
{
"PCA1": users_2[:, 0],
"PCA2": users_2[:, 1],
"labels": labels
},
index=df_freq.index)
data = [
go.Scatter(x=activity_2[activity_2["labels"] == label]["PCA1"],
y=activity_2[activity_2["labels"] == label]["PCA2"],
mode="markers",
text=df_freq.index,
name=str(label)) for label in activity_2["labels"].unique()
]
py.plot(data)
def display_corr_activities(frequency_path="data/data_frequency.json"):
"""
Display the correlation matrix of activities described by users' use-frequency
"""
df = dataframe_activity_frequency()
corr = df.corr()
# corr.style.background_gradient(cmap="coolwarm",axis=None)
data = [go.Heatmap(z=corr, x=corr.columns, y=corr.index)]
py.plot(data)
def display_corr_principal_users(frequency_path="data/data_frequency.json"):
"""
Display the correlation matrix of users described by activities' use-frequency
"""
df = dataframe_activity_frequency().transpose()
corr = df.corr()
# corr.style.background_gradient(cmap="coolwarm",axis=None)
data = [go.Heatmap(z=corr, x=corr.columns, y=corr.index)]
layout = go.Layout(xaxis={"showticklabels": False},
yaxis={"showticklabels": False},
width=700,
height=700)
py.plot({"data": data, "layout": layout})
def pca_kmeans_users(n_clusters=3,
frequency_filepath="data/data_frequency.json"):
"""
User observations with PCA described by use-frequency of activities
Compute K-means algorithm
:param n_clusters int (optional): Number of clusters for K-means algorithm
"""
df_freq = dataframe_activity_frequency()
# print(df_freq.head())
pca = PCA(n_components=2)
pca.fit(df_freq)
# print(pca.components_)
# print(pca.explained_variance_ratio_)
users_2 = pca.transform(df_freq)
# data = [go.Scatter(
# x=users_2[:,0],
# y=users_2[:,1],
# mode = "markers"
# )]
kmeans = KMeans(n_clusters=n_clusters).fit(df_freq)
labels = kmeans.labels_
activity_2 = pd.DataFrame(
{
"PCA1": users_2[:, 0],
"PCA2": users_2[:, 1],
"labels": labels
},
index=df_freq.index)
data = [
go.Scatter(x=activity_2[activity_2["labels"] == label]["PCA1"],
y=activity_2[activity_2["labels"] == label]["PCA2"],
mode="markers",
text=df_freq.index,
name=str(label),
marker={"size": 15})
for label in np.sort(activity_2["labels"].unique())
]
layout = go.Layout(
xaxis={
"title":
"PC1 {}% of explained variance".format(
str(pca.explained_variance_ratio_[0])[2:4])
},
yaxis={
"title":
"PC2 {}%".format(str(pca.explained_variance_ratio_[1])[2:4])
},
)
py.plot({"data": data, "layout": layout})
def pca_kmeans_activities(n_clusters=3,
therapy_filepath="data/therapyByUser.json"):
"""
Activity observations with PCA described by users' frequency
Compute K-means algorithm
:param n_clusters int (optional): Number of clusters for K-means algorithm
"""
df = dataframe_activity_frequency()
df = df.transpose()
# print(df)
# print(df.describe())
pca = PCA(n_components=2)
pca.fit(df)
# print(pca.components_)
# print(pca.explained_variance_ratio_)
activity_2 = pca.transform(df)
kmeans = KMeans(n_clusters=n_clusters).fit(activity_2)
labels = kmeans.labels_
activity_2 = pd.DataFrame(
{
"PCA1": activity_2[:, 0],
"PCA2": activity_2[:, 1],
"labels": labels
},
index=df.index)
# print(activity_2)
data = [
go.Scatter(x=activity_2[activity_2["labels"] == label]["PCA1"],
y=activity_2[activity_2["labels"] == label]["PCA2"],
mode="markers",
text=activity_2[activity_2["labels"] == label].index,
name=str(label),
marker={"size": 15})
for label in activity_2["labels"].unique()
]
layout = go.Layout(xaxis={
"title":
"PC1 {}% of explained variance".format(
str(pca.explained_variance_ratio_[0])[2:4])
},
yaxis={
"title":
"PC2 {}%".format(
str(pca.explained_variance_ratio_[1])[2:4])
})
py.plot({"data": data, "layout": layout})