I have data that represent a time series of categorical variables. I want to display the transitions in categories below a traditional line plot of related continuous time series to show off context as time evolves. I'd like to know the best way to do this. My attempt was in terms of Rectangles. The appearance is a bit weird, and importantly the axis labels for the x axis don't render as dates.
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
from pandas.plotting import register_matplotlib_converters
import matplotlib.dates as mdates
register_matplotlib_converters()
t0 = pd.DatetimeIndex(["2017-06-01 00:00","2017-06-17 00:00","2017-07-03 00:00","2017-08-02 00:00","2017-08-09 00:00","2017-09-01 00:00"])
t1 = pd.DatetimeIndex(["2017-06-01 00:00","2017-08-15 00:00","2017-09-01 00:00"])
df0 = pd.DataFrame({"cat":[0,2,1,2,0,1]},index = t0)
df1 = pd.DataFrame({"op":[0,1,0]},index=t1)
# Create new plot
fig,ax = plt.subplots(1,figsize=(8,3))
data_layout = {
"cat" : {0: ('bisque','Low'),
1: ('lightseagreen','Medium'),
2: ('rebeccapurple','High')},
"op" : {0: ('darkturquoise','Open'),
1: ('tomato','Close')}
}
vars =("cat","op")
dfs = [df0,df1]
all_ticks = []
leg = []
for j,(v,d) in enumerate(zip(vars,dfs)):
dvals = d[v][:].astype("d")
normal = mpl.colors.Normalize(vmin=0, vmax=2.)
colors = plt.cm.Set1(0.75*normal(dvals.as_matrix()))
handles = []
for i in range(d.count()-1):
s = d[v].index.to_pydatetime()
level = d[v][i]
base = d[v].index[i]
w = s[i+1] - s[i]
patch=mpl.patches.Rectangle((base,float(j)),width=w,color=data_layout[v][level][0],height=1,fill=True)
ax.add_patch(patch)
for lev in data_layout[v]:
print data_layout[v][level]
handles.append(mpl.patches.Patch(color=data_layout[v][lev][0],label=data_layout[v][lev][1]))
all_ticks.append(j+0.5)
leg.append( plt.legend(handles=handles,loc = (3-3*j+1)))
plt.axhline(y=1.,linewidth=3,color="gray")
plt.xlim(pd.Timestamp(2017,6,1).to_pydatetime(),pd.Timestamp(2017,9,1).to_pydatetime())
plt.ylim(0,2)
ax.add_artist(leg[0]) # two legends on one axis
ax.format_xdata = mdates.DateFormatter('%Y-%m-%d') # This fails
plt.yticks(all_ticks,vars)
plt.show()
which produces this with no dates and has jittery lines:. How do I fix this? Is there a better way entirely?
This is a way to display dates on x-axis:
In your code substitute the line that fails with this one:
ax.xaxis.set_major_formatter((mdates.DateFormatter('%Y-%m-%d')))
But I don't remember how it should look like, can you show us the end-result again?
Related
I have a Pandas database (below I am creating a random database to mimic mine).
I use Seaborn, facet-grid and scatterplot to plot the data the way I want : Epsilon1 as a function of no, I distinguish the data from the sub categories A and B using different subplots and colors. This part of the code works correctly.
Then I want that the user can click on any dot in order to display in the IPython console and in the status bar of the Matplotlib figure (as here) all the informations about this dot : that is to say all the values of the corresponding dataframe row: something like:
'no':5, 'Date':1997-12-15 03:50:41, 'A':A6, 'B':B4, 'Epsilon1':0.670635, 'Epsilon2':0.756461, 'Epsilon3':0.530825
I have made first tests using onpick event (not shown here) but all were unsuccessful.
Actually I can't get by with the function onpick(event) because I do not understand why print(event.ind) gives me a list of integers...
Here is my code
import pandas as pd
import numpy as np
import seaborn as sns
import random
# size of the database
n = 1000
nA = 6
nB = 5
no = np.arange(n)
date = np.random.randint(1e9, size=n).astype('datetime64[s]')
A = [''.join(['A',str(random.randint(1, nA))]) for j in range(n)]
B = [''.join(['B',str(random.randint(1, nB))]) for j in range(n)]
Epsilon1 = np.random.random_sample((n,))
Epsilon2 = np.random.random_sample((n,))
Epsilon3 = np.random.random_sample((n,))
data = pd.DataFrame({'no':no,
'Date':date,
'A':A,
'B':B,
'Epsilon1':Epsilon1,
'Epsilon2':Epsilon2,
'Epsilon3':Epsilon3})
def onpick(event):
print(event.ind)
def plot_Epsilon1_seaborn():
sns.set_theme()
g = sns.FacetGrid(data,
col="A",
col_wrap=4,
hue='B',
hue_order=data['B'].sort_values().drop_duplicates().to_list(),
palette="viridis",
col_order=data['A'].sort_values().drop_duplicates().to_list())
g.map(sns.scatterplot,
'no',
'Epsilon1',
picker=True)
g.add_legend()
g.fig.canvas.mpl_connect("pick_event", onpick)
if __name__ == '__main__':
plot_Epsilon1_seaborn()
Am trying to find hist()'s figsize and layout parameter for sns.pairplot().
I have a pairplot that gives me nice scatterplots between the X's and y. However, it is oriented horizontally and there is no equivalent layout parameter to make them vertical to my knowledge. 4 plots per row would be great.
This is my current sns.pairplot():
sns.pairplot(X_train,
x_vars = X_train.select_dtypes(exclude=['object']).columns,
y_vars = ["SalePrice"])
This is what I would like it to look like: Source
num_mask = train_df.dtypes != object
num_cols = train_df.loc[:, num_mask[num_mask == True].keys()]
num_cols.hist(figsize = (30,15), layout = (4,10))
plt.show()
What you want to achieve isn't currently supported by sns.pairplot, but you can use one of the other figure-level functions (sns.displot, sns.catplot, ...). sns.lmplot creates a grid of scatter plots. For this to work, the dataframe needs to be in "long form".
Here is a simple example. sns.lmplot has parameters to leave out the regression line (fit_reg=False), to set the height of the individual subplots (height=...), to set its aspect ratio (aspect=..., where the subplot width will be height times aspect ratio), and many more. If all y ranges are similar, you can use the default sharey=True.
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
# create some test data with different y-ranges
np.random.seed(20230209)
X_train = pd.DataFrame({"".join(np.random.choice([*'uvwxyz'], np.random.randint(3, 8))):
np.random.randn(100).cumsum() + np.random.randint(100, 1000) for _ in range(10)})
X_train['SalePrice'] = np.random.randint(10000, 100000, 100)
# convert the dataframe to long form
# 'SalePrice' will get excluded automatically via `melt`
compare_columns = X_train.select_dtypes(exclude=['object']).columns
long_df = X_train.melt(id_vars='SalePrice', value_vars=compare_columns)
# create a grid of scatter plots
g = sns.lmplot(data=long_df, x='SalePrice', y='value', col='variable', col_wrap=4, sharey=False)
g.set(ylabel='')
plt.show()
Here is another example, with histograms of the mpg dataset:
import matplotlib.pyplot as plt
import seaborn as sns
mpg = sns.load_dataset('mpg')
compare_columns = mpg.select_dtypes(exclude=['object']).columns
mpg_long = mpg.melt(value_vars=compare_columns)
g = sns.displot(data=mpg_long, kde=True, x='value', common_bins=False, col='variable', col_wrap=4, color='crimson',
facet_kws={'sharex': False, 'sharey': False})
g.set(xlabel='')
plt.show()
I have the following dataset, ratings in stars for two fictitious places:
import pandas as pd
import matplotlib.pyplot as plt
df = pd.DataFrame({'id':['A','A','A','A','A','A','A','B','B','B','B','B','B'],
'rating':[1,2,4,5,5,5,3,1,3,3,3,5,2]})
Since the rating is a category (is not a continuous data) I convert it to a category:
df['rating_cat'] = pd.Categorical(df['rating'])
What I want is to create a bar plot per each fictitious place ('A or B'), and the count per each rating. This is the intended plot:
I guess using a for per each value in id could work, but I have some trouble to decide the size:
fig, ax = plt.subplots(1,2,figsize=(6,6))
axs = ax.flatten()
cats = df['rating_cat'].cat.categories.tolist()
ids_uniques = df.id.unique()
for i in range(len(ids_uniques)):
ax[i].bar(df[df['id']==ids_uniques[i]], df['rating'].size())
But it returns me an error TypeError: 'int' object is not callable
Perhaps it's something complicated what I am doing, please, could you guide me with this code
The pure matplotlib way:
from math import ceil
# Prepare the data for plotting
df_plot = df.groupby(["id", "rating"]).size()
unique_ids = df_plot.index.get_level_values("id").unique()
# Calculate the grid spec. This will be a n x 2 grid
# to fit one chart by id
ncols = 2
nrows = ceil(len(unique_ids) / ncols)
fig = plt.figure(figsize=(6,6))
for i, id_ in enumerate(unique_ids):
# In a figure grid spanning nrows x ncols, plot into the
# axes at position i + 1
ax = fig.add_subplot(nrows, ncols, i+1)
df_plot.xs(id_).plot(axes=ax, kind="bar")
You can simplify things a lot with Seaborn:
import seaborn as sns
sns.catplot(data=df, x="rating", col="id", col_wrap=2, kind="count")
If you're ok with installing a new library, seaborn has a very helpful countplot. Seaborn uses matplotlib under the hood and makes certain plots easier.
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
df = pd.DataFrame({'id':['A','A','A','A','A','A','A','B','B','B','B','B','B'],
'rating':[1,2,4,5,5,5,3,1,3,3,3,5,2]})
sns.countplot(
data = df,
x = 'rating',
hue = 'id',
)
plt.show()
plt.close()
Even though there are nice examples online about plotting side by side boxplots. With the way my data is set in two different pandas DataFrames and allready having sum subplots I have not been able to manage getting my boxplots next to each other in stead of overlapping.
my code is as follows:
import matplotlib as mpl
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
mpl.use('agg')
fig, axarr = plt.subplots(3,sharex=True,sharey=True,figsize=(9,6))
month = ['jan','feb','mar','apr','may','jun','jul','aug','sep','oct','nov','dec']
percentiles = [90,95,98]
nr = 0
for p in percentiles:
future_data = pd.DataFrame(np.random.randint(0,30,size=(30,12)),columns = month)
present_data = pd.DataFrame(np.random.randint(0,30,size=(30,12)),columns = month)
Future = future_data.as_matrix()
Present = present_data.as_matrix()
pp = axarr[nr].boxplot(Present,patch_artist=True, showfliers=False)
fp = axarr[nr].boxplot(Future, patch_artist=True, showfliers=False)
nr += 1
The results looks as follows:
Overlapping Boxplots
Could you help me out in how to makes sure the boxes are next to each other so I can compare them without being bothered by the overlap?
Thank you!
EDIT: I have reduced the code somewhat so it can run like this.
You need to position your bars manually, i.e. providing the positions as array to the position argument of boxplot. Here it makes sense to shift one by -0.2 and the other by +0.2 to their integer position. You can then adjust the width of them to sum up to something smaller than the difference in positions.
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
fig, axarr = plt.subplots(3,sharex=True,sharey=True,figsize=(9,6))
month = ['jan','feb','mar','apr','may','jun','jul','aug','sep','oct','nov','dec']
percentiles = [90,95,98]
nr = 0
for p in percentiles:
future_data = pd.DataFrame(np.random.randint(0,30,size=(30,12)),columns = month)
present_data = pd.DataFrame(np.random.randint(0,30,size=(30,12)),columns = month)
Future = future_data.as_matrix()
Present = present_data.as_matrix()
pp = axarr[nr].boxplot(Present,patch_artist=True, showfliers=False,
positions=np.arange(Present.shape[1])-.2, widths=0.4)
fp = axarr[nr].boxplot(Future, patch_artist=True, showfliers=False,
positions=np.arange(Present.shape[1])+.2, widths=0.4)
nr += 1
axarr[-1].set_xticks(np.arange(len(month)))
axarr[-1].set_xticklabels(month)
axarr[-1].set_xlim(-0.5,len(month)-.5)
plt.show()
I am trying to produce a visualization of the SOM mapping for the Iris dataset ( https://archive.ics.uci.edu/ml/datasets/Iris).
My code so far:
from sklearn.datasets import load_iris
from mvpa2.suite import *
import pandas as pd
import numpy as np
df = pd.read_csv(filepath_or_buffer='data/iris.data', header=None, sep=',')
df.columns=['sepal_len', 'sepal_wid', 'petal_len', 'petal_wid', 'class']
df.dropna(how="all", inplace=True) # drops the empty line at file-end
# split the data table into feature data x and class labels y
x = df.ix[:,0:4].values # the first 4 columns are the features
y = df.ix[:,4].values # the last column is the class label
t = np.zeros(len(y), dtype=int)
t[y == 'Iris-setosa'] = 0
t[y == 'Iris-versicolor'] = 1
t[y == 'Iris-virginica'] = 2
som = SimpleSOMMapper((240, 320), 100, learning_rate=0.05)
som.train(x)
pl.imshow(som.K, origin='lower')
mapped = som(x)
for i, m in enumerate(mapped):
pl.text(m[1], m[0], t[i], ha='center', va='center',
bbox=dict(facecolor='white', alpha=0.5, lw=0))
pl.show()
which produces this mapping:
Is there any way to customize the palette so it looks nicer like this one? (taken from https://github.com/JustGlowing/minisom)?
Basically I am trying to use a nicer palette (perhaps with fewer colors) and mark the class labels in a nicer way.
Thank you.
I will answer my own question: it turns out that I forgot to slice my data:
pl.imshow(som.K[:,:,0], origin='lower')
Everything looks fine now: