seaborn

Easy Create Mosaic Plot using Stacked Bar Chart

Creating Mosaic Plot

In one of my work project, I need to use mosaic plot to visualize the proportion of different variables/elements exists in each group.  It is hard to find a readily available mosaic plot function (from Seaborn etc) which can be easily customized. By reading some of the blogs, mosaic plot can be created using stacked bar chart concept by performing some transformation on the raw data and overlaying individual bar charts. With this knowledge and using python Pandas and Matplotlib, I am able to create a mosaic plot that is good enough for my need.

Sample Data Sets

A sample data set is as shown below. We need to plot the proportion of b, g, r (all the columns) for each index (0 to 4). Based on the format of the data set, we make a transformation of the columns to be able to have Mosaic Plot.

Breaking down the data transformation for stacked bar chart plotting

We perform two transformations as followed. Mosaic plot requires the sum of  proportion of categories for each group to be 1.0 or 100%. Stacked bar chart can achieve this by summing or stacking values for each element in the group but we would need to ensure the values are normalized and the sum of all elements in a group equal to 1 (i.e r+ g+b =1 for each index).

To simulate the effect of stacked bar chart , the trick is to use multiple bar charts to overlay on top of each other to simulate the effect of stacked bar chart. To be able to create the stacked effect, the ratio/proportion of the stacked element need to be the sum of proportion value of “bottom” elements + the proportion value of the element itself. This can be easily achieved by doing a cumulative sum along the row axis.

As example below, r will be used as a base (since values are based on b + g + r). g will overlay on top of r since it is summation of b + g. b will be final layer overlay on g and r.

Mosaic plot function

Once the transformations are done, it is easy to plot the mosaic plot by plotting the different bar charts and overlaying on top of each other. Additional module adjustText can be used to prevent overlapping of the text labels in the plot. Based on the above, we can create a general mosaic function as below.

 

Advertisements

Monitoring quality over time with heap map

A particular concern with testing hard disk drives over multiple times is the quality of certain drives may degrade (wear and tear) over time and we failed to detect this degradation.

We have certain metrics to gauge any degradation symptom observed for a particular head in a particular drive. For example, with metric A, we are looking at the % change over time reference to the date of the first test o determine whether a head is degraded.

Below python code will base on the following table to generate the required heatmap for easy visualization.

untitled

Calculating %Change

import seaborn as sns
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

df1['DATE1'] = df1.DATE.dt.strftime('%m/%d/%Y')
df1 = df1.sort_values(by = 'DATE1')

# calculate the metric % change and
# actual change with reference to each individual head first data

df1['METRIC_A_PCT_CHANGE'] = df1.groupby(['SERIAL','HEAD'])['METRIC_A']\
                            .apply(lambda x: x.div(x.iloc[0]).subtract(1).mul(100))
df1['METRIC_A_CHANGE'] = df1.groupby(['SERIAL','HEAD'])['METRIC_A']\
                         .apply(lambda x: x - x.iloc[0])

Plotting in HeapMap

fig, ax = plt.subplots(figsize=(10,10))

# Pivot it for plotting in heap map
ww = df1.pivot_table(index = ['SERIAL','HEAD'], \
                     columns = 'DATE1', values = "METRIC_A_PCT_CHANGE")

g = sns.heatmap(ww, vmin= -5, vmax = 5, center = 0, \
                cmap= sns.diverging_palette(220, 20, sep=20, as_cmap=True),\
                xticklabels=True, yticklabels=True, \
                ax = ax, linecolor = 'white', linewidths = 0.1, annot = True)

g.set_title("% METRIC_A changes over multiple Dates", \
            fontsize = 16, color = 'blue')

 

Generated Plots

From the heap map, SER_3BZ-0 have some indication of degradation with increasing % Metric A loss over the different test date.

untitled

Notes

  • Getting the % percentage change relative to first value of each group.
    • df.groupby(‘security’)[‘price’].apply(lambda x: x.div(x.iloc[0]).subtract(1).mul(100))

 

Custom Contour Plots with Labelled points

Creating Customized Contour Plots with Labelled Points

I was asked to create a customized contour plot based on a chart (Fig 1 ) found in IEEE Transactions on Magnetics journal with some variant in requirements. The chart shows the areal density capacity (ADC) demo of certain samples on a bit density (BPI) by track density (TPI) chart. The two different contours shown in the plot are made up of ADC (BPI * TPI) and bit aspect ratio BAR (BPI/TPI).

A way to create the plot might be to generate the contours based on Excel and manually added in the different points. This proves to be too much work. Therefore, a simpler way is needed. Further requirements include having additional points (with labels) to be added in fairly easily and charts with different sets of data can be recreated rapidly.

Creating the Contours

The idea will be to use the regression plots for both the ADC and the BAR contours while the points and labels can be automatically added to the plots after reading from an Excel table (or csv file). The regression plots are based on seaborn lmplot and the points with labels are annotated on the chart based on the individual x, and y values.

Besides the seaborn, pandas, matplotlib and numpy,  additional module adjustText is used to prevent overlapping of the text labels in the plot

import seaborn as sns
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from adjustText import adjust_text

## Create GridLines for the ADC GBPSI
ADC_tgt = range(650,2150,50)
BPI_tgt = list(range(800,2700,20))*3
data_list = [ [ADC, BPI, ADC*1000/BPI] for BPI in BPI_tgt for ADC in ADC_tgt]
ADC_df = pd.DataFrame(data_list, columns=['Contour','X','Y']) #['ADC','TPI','BPI']
ADC_df['Contour'] = ADC_df['Contour'].astype('category')

## Create GridLines for the BAR
BAR_tgt =[1.0,1.5,2.0, 2.5,3.0,3.5,4.0,4.5,5.0,5.5,6.0,6.5]
BPI_tgt = list(range(800,2700,20))*3
data_list = [ [BAR, BPI, BPI/BAR] for BPI in BPI_tgt for BAR in BAR_tgt]
BAR_df = pd.DataFrame(data_list, columns=['Contour','X','Y']) #['BAR','TPI','BPI']
BAR_df['Contour'] = BAR_df['Contour'].astype('category')

combined_df = pd.concat([ADC_df,BAR_df])

Adding the demo points with text from Excel

The various points are updated in the excel sheet (or csv) , shown in fig 2, and read using pandas. Two data frames are produced, pts_df and text_df which is the dataframe from the points and the associated text. These, together with the contour data frame from above, are then feed into the seaborn lmplot. Note the points shown in the Excel and plots are randomly generated.

class ADC_DataPts():

    def __init__(self, xls_fname, header_psn = 0):
        self.xls_fname = xls_fname
        self.header_psn = header_psn
        self.data_df = pd.read_excel(self.xls_fname, header = self.header_psn)

    def generate_pts_text_df(self):
        pts_df = self.data_df['X Y Color'.split()]
        text_df = self.data_df['X_TxtPsn Y_TxtPsn TextContent'.split()]
        return pts_df, text_df

data_excel = r"yourexcelpath.xls"
adc_data = ADC_DataPts(data_excel, header_psn =1)
pts_df, text_df = adc_data.generate_pts_text_df()

Seaborn lmplot

The seaborn lmplot is used for the contours while the points are individually annotated on the graph

def generate_contour_plots_with_points(xlabel, ylabel, title):

    # overall settings for plots
    sns.set_context("talk")
    sns.set_style("whitegrid", \
                  {'grid.linestyle': ':', 'xtick.bottom': True, 'xtick.direction': 'out',\
                    'xtick.color': '.15','axes.grid' : False}
                 )

    # Generate the different "contour"
    g = sns.lmplot("X", "Y", data=combined_df, hue='Contour', order =2, \
               height =7, aspect =1.5, ci =False, line_kws={'color':'0.9', 'linestyle':':'}, \
                scatter=False, legend_out =False)

    # Bold the key contour lines
    for n in [1.0,2.0,3.0]:
        sub_bar = BAR_df[BAR_df['Contour']==n]
        #generate the bar contour
        g.map(sns.regplot, x= "X", y="Y", data=sub_bar ,scatter= False, ci =False, \
              line_kws={'color':'0.9', 'linestyle':'-', 'alpha':0.05, 'linewidth':'3'})

    for n in [1000,1500,2000]:
        sub_adc = ADC_df[ADC_df['Contour']==n]
        #generate the bar contour
        g.map(sns.regplot, x= "X", y="Y", data=sub_adc ,scatter= False, ci =False, order =2, \
              line_kws={'color':'0.9', 'linestyle':'-', 'alpha':0.05, 'linewidth':'3'})#'color':'0.7', 'linestyle':'-', 'alpha':0.05, 'linewidth':'2'

    # Generate the different points
    for index, rows in pts_df.iterrows():
        g = g.map_dataframe(plt.plot, rows['X'], rows['Y'], 'o',  color = rows['Color'])# generate plot with differnt color or use annotation?

    ax = g.axes.flat[0]    

    # text annotation on points
    style = dict(size=12, color='black', verticalalignment='top')
    txt_grp = []
    for index, rows in text_df.iterrows():
        txt_grp.append(ax.text( rows['X_TxtPsn'], rows['Y_TxtPsn'], rows['TextContent'], **style) )#how to find space, separate data base

    style2 = dict(size=12, color='grey', verticalalignment='top')
    style3 = dict(size=12, color='grey', verticalalignment='top', rotation=30, alpha= 0.7)

    # Label the key contours
    ax.text( 2400, 430, '1000 Gfpsi', **style2)
    ax.text( 2400, 640, '1500 Gfpsi', **style2)
    ax.text( 2400, 840, '2000 Gfpsi', **style2) 

    ax.text( 1100, 570, 'BAR 2.0', **style3)
    ax.text( 1300, 460, 'BAR 3.0', **style3) 

    # Set x y limit
    ax.set_ylim(400,1000)
    ax.set_xlim(1000,2600)

    # Set general plot attributes
    g.set_xlabels(xlabel)
    g.set_ylabels(ylabel)
    plt.title(title)

    adjust_text(txt_grp, x = pts_df.X.tolist() , y = pts_df.Y.tolist() , autoalign = True, expand_points=(1.4, 1.4))

generate_contour_plots_with_points('kBPI', 'kTPI', "DEMO Areal Density Capability\n")
Untitled

Fig 1: Sample plot from Heat-Assisted Interlaced Magnetic Recording IEEE Vol 54 No2

Untitled

Fig2: Excel tables with associated demo points, the respective color and the text labels

Untitled

Fig 3: Generated chart with the ADC and BAR contours and demo pts with labels

Heap Map for discrepancy check

Monitoring counts discrepancy

In one aspect of my work, we have a group of samples undergoing several rounds of modifications with same set of tests being performed at each round. For each test, parameters for each sample are collected. For some samples, a particular test may fail in certain rounds resulting in no/missing parameters being collected for that test.

When we compare the performance of the samples especially grouping as a mean, missing parameters from certain samples at certain rounds may skew the results. To ensure accuracy, we need to ensure matching samples data. As there are multiple tests and few hundreds parameters being tracked, we need a way to keep track of the parameters that have mismatch parameters between rounds.

A simple way will be to use the heat map to highlight parameters that have discrepancy in number of counts (this will mean that some samples are missing in data) between rounds. The script is generated using mainly Pandas and Seaborn.

Steps

  1. Group the counts for each parameter for each round.
  2. Use one round as reference (default 1st round), take the differences in counts for each parameter for each round.
  3. Display as heat map for only rounds that have discrepancy.
import os, sys, datetime, re
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

# retrieve zone data
rawfile = 'raw_data.csv'
raw_df = pd.read_csv(rawfile)

# count of data in group
cnt_df = raw_df.groupby(['round']).count()

# Substract the first to the rest
diff_df = cnt_df.subtract(cnt_df.iloc[0], axis = 1)

# drop columns where it is all zeros, meaning exclude data that are matched.
diff_df.loc[:, diff_df.any()]

fig, ax = plt.subplots(figsize=(10,10))  

sns.heatmap(diff_df.loc[:, diff_df.any()].T,  xticklabels=True, yticklabels=True, ax =ax , annot=True, fmt="d", center= 0 ,  cmap="coolwarm")
plt.tight_layout()

Untitled

Extra

Quick view of missing data using seaborn heatmap


sns.heatmap(df.isnull(), yticklabels=False, cbar = False, cmap = 'viridis')

missingdata