APIs de Visualização em Python
8 likes8,825 views
The document discusses various Python libraries for data visualization, including Seaborn, Bokeh, and Vincent, emphasizing their integration with Pandas and Numpy for exploratory data analysis. It details the visualization techniques applied to a dataset on the top 100 wines and provides examples of graphical representations using ggplot and Bokeh for stock data analysis. The author also references several online resources and libraries for enhancing the visualization experience.
![Qual a relação entre Pontos e Prêmios por País?
idx = (wty.Country == 'AUSTRALIA') |
(wty.Country == 'FRANCE') |
(wty.Country == 'SPAIN') |
(wty.Country == 'USA') |
(wty.Country == 'ARGENTINA')
gplot(wty[idx], aes(x="Prizes", y="Points")) +
geom_point() +
stat_smooth(method='lm', color='blue') +
facet_wrap("Country", scales="fixed")](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-16-320.jpg)
![Como se distribuem os prêmios por País?
wty_country_prizes = wty['Prizes'].groupby(wty['Country'])
idx = wty_country_prizes.count() == 1
wty_t = wty.ix[[c not in idx[idx].index for c in wty.Country],
['Country','Prizes']]
wty_t_country_mean = wty_t.groupby(wty_t.Country).mean()
country_names = wty_t_country_mean.Prizes.sort(inplace=False).index
import seaborn as sb
sb.set(style="ticks")
f, ax = plt.subplots()
sb.offset_spines()
sb.violinplot(wty_t['Prizes'], wty_t['Country'], order=country_names)
locs, labels = plt.xticks()
plt.setp(labels, rotation=45)
sb.despine(trim=True)](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-17-320.jpg)
![wty_country = wty.groupby('Country')
country_count = wty_country['Wine'].aggregate(len)
country_count.sort(inplace=True)
fig, ax = plt.subplots(figsize=(12,8))
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.get_xaxis().tick_bottom()
ax.get_yaxis().tick_left()
plt.subplot(221)
g = wty_country['Prizes'].count()
x = np.arange(len(g))
countries = list(g.index.values)
y = g.values
plt.bar(x+1, y, align='center', width=0.5, alpha=0.4)
plt.xticks(x+1, [' ']*10, rotation=90)
plt.title('Contagem')
plt.ylabel('Prizes')
plt.subplot(222)
g = wty_country['Prizes'].mean()
y = g.values
plt.bar(x+1, y, align='center', color='red', width=0.5, alpha=0.4)
plt.xticks(x+1, [' ']*10, rotation=90);
plt.title(u'Média')
plt.subplot(223)
g = wty_country['Prizes'].max()
y = g.values
plt.bar(x+1, y, align='center', color='darkgreen', width=0.5, alpha=0.4)
plt.xticks(x+1, countries, rotation=90);
plt.title(u'Máx')
plt.ylabel('Prizes')
plt.subplot(224)
g = wty_country['Prizes'].min()
y = g.values
plt.bar(x+1, y, align='center', color='magenta', width=0.5, alpha=0.4)
plt.xticks(x+1, countries, rotation=90);
plt.title(u'Mín')
Mas qual País possui o melhor vinho?](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-18-320.jpg)
![>>> mpld3.save_html(fig, 'fig.html')
<div id="fig_el672345700473765832459779"></div>
<script>
function mpld3_load_lib(url, callback){
var s = document.createElement('script');
s.src = url;
s.async = true;
s.onreadystatechange = s.onload = callback;
s.onerror = function(){console.warn("failed to load library " + url);};
document.getElementsByTagName("head")[0].appendChild(s);
}
if(typeof(mpld3) !== "undefined" && mpld3._mpld3IsLoaded){
// already loaded: just create the figure
!function(mpld3){
...
mpld3 salva fig em html](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-19-320.jpg)
![import Quandl
tickers = [
"GOOG/BVMF_BOVA11.4",
"YAHOO/SA_PETR3.6",
"GOOG/BVMF_VALE5.4",
"GOOG/BVMF_BBDC4.4",
"GOOG/BVMF_BBAS3.4",
"GOOG/BVMF_ITUB4.4"]
series = Quandl.get(tickers,
trim_start="2013-01-01",
trim_end="2013-12-31")
series.columns = [
"BOVESPA",
"PETROBRAS",
"VALE",
"BRADESCO",
"BB",
"ITAU"]
series.dropna(inplace=True)
dados: Ações (Quandl)
>>> print series
BOVESPA PETROBRAS VALE BRADESCO BB ITAU
Date
2013-01-02 61.27 19.00 42.60 36.02 25.80 31.3
2013-01-03 61.92 19.70 42.09 38.12 26.31 32.2
2013-01-04 61.32 19.77 41.17 37.45 26.00 31.9
2013-01-07 60.75 19.51 40.35 37.29 26.15 31.5
2013-01-08 59.91 18.96 40.10 37.42 26.45 31.6
2013-01-09 60.25 19.15 40.30 37.90 26.70 32.1
2013-01-10 60.49 19.27 40.57 37.40 26.30 31.7
2013-01-11 60.30 19.32 39.53 37.27 25.98 31.8
2013-01-14 60.75 19.21 39.77 37.88 26.93 32.2
2013-01-15 60.52 19.27 39.65 37.30 26.70 31.8
2013-01-16 60.54 19.34 39.68 37.49 26.86 32.1
2013-01-17 61.01 19.21 39.75 38.11 26.81 32.6
2013-01-18 60.79 19.05 39.20 38.25 26.20 32.7
2013-01-21 60.75 18.87 39.15 38.09 25.93 32.5
2013-01-22 60.55 19.12 39.59 38.40 25.64 32.9
...](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-20-320.jpg)
![Exporta a Gramática de Gráficos
>>> print line.to_json()
{
"axes": [
{ "scale": "x", "title": "Date", "type": "x" },
{ "scale": "y", "title": "Price", "type": "y" }
],
"data": [
{
"name": "table",
"values": [
{ "col": "BOVESPA", "idx": 1357092000000, "val": 61.27 },
{ "col": "PETROBRAS","idx": 1357092000000, "val": 19.0 },
{ "col": "VALE", "idx": 1357092000000, "val": 42.6 },
...](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-22-320.jpg)
![from bokeh.plotting import *
output_notebook()
import numpy as np
import Quandl
import pandas as pd
from collections import OrderedDict
tickers = ["GOOG/BVMF_BOVA11.4", "GOOG/BVMF_PETR3.4", "GOOG/BVMF_VALE5.4", 'GOOG/BVMF_BBDC4.4', 'GOOG/BVMF_BBAS3.4', 'GOOG/BVMF_ITUB4.4', 'GOOG/BVMF_USIM5.4', 'GOOG/BVMF_CIEL3.
4', 'GOOG/BVMF_CMIG4.4', 'GOOG/BVMF_CTIP3.4', 'GOOG/BVMF_CPLE6.4', 'GOOG/BVMF_ELET6.4', 'GOOG/BVMF_GGBR4.4', 'GOOG/BVMF_GOLL4.4', 'GOOG/BVMF_HGTX3.4', 'GOOG/BVMF_JBSS3.4',
'GOOG/BVMF_MRFG3.4', 'GOOG/BVMF_PDGR3.4']
series = Quandl.get(tickers, trim_start="2013-01-01", trim_end="2013-12-31", authtoken="nJ1NhTYdEs2p3MsS4CVd")
series.columns = ['BOVESPA', 'PETROBRAS', 'VALE', 'BRADESCO', 'BB', 'ITAU', 'USIMINAS', 'CIELO', 'CEMIG', 'CETIP', 'COPEL', 'ELETROPAULO', 'GERDAU', 'GOL', 'HERING', 'JBS',
'MAFRIG', 'PDGR']
series.dropna(inplace=True)
logreturns = np.log(series).diff().dropna()
months = ["Jan", "Fev", "Mar", "Abr", "Mai", "Jun", "Jul", "Ago", "Set", "Out", "Nov", "Dez"]
series_mr = logreturns.resample('M', how=sum)*100
series_mr['Month'] = [months[d.month-1] for d in series_mr.index]
series_mr.set_index('Month', inplace=True)
months = list(series_mr.index)
stocks = list(series_mr.columns)
import brewer2mpl
bmap = brewer2mpl.get_map('RdYlGn', 'diverging', 3)
colors = bmap.hex_colors
levels = np.array([0, 1])
month = []
stock = []
color = []
rate = []
for y in stocks:
for m in months:
month.append(m)
stock.append(y)
monthly_rate = series_mr[y][m]
rate.append(monthly_rate)
color.append(colors[sum(levels < monthly_rate)])
source = ColumnDataSource(data=dict(month=month, year=stock, color=color, rate=rate, ))
figure()
rect('year', 'month', 0.95, 0.95, source=source, x_range=stocks, y_range=list(reversed(months)), color='color', line_color=None, tools="resize,hover,save", title="Rentabilidade
de ações", plot_width=300, plot_height=400)
grid().grid_line_color = None
axis().axis_line_color = None
axis().major_tick_line_color = None
axis().major_label_text_font_size = "5pt"
axis().major_label_standoff = 0
xaxis().location = "top"
xaxis().major_label_orientation = np.pi/3
from bokeh.objects import HoverTool
hover = [t for t in curplot().tools if isinstance(t, HoverTool)][0]
hover.tooltips = OrderedDict([('date', '@month'), ('rate', '@rate'),])
show()
Rentabilidade de ações
Esse gráfico já não é tão simples!
;-)](https://image.slidesharecdn.com/apisdevisualizacaoempython-140809203923-phpapp01/85/APIs-de-Visualizacao-em-Python-29-320.jpg)
APIs de Visualização em Python
- 1. APIs de Visualização em Python TDC 2014 Wilson Freitas http://aboutwilson.net @aboutwilson
- 2. Visualização
- 5. ggplot
- 6. seaborn ● Biblioteca de visualização baseada no matplotlib " seaborn is a library for making attractive and informative statistical graphics in Python " ● Boa para análise exploratória de dados ● Fortemente integrada com pandas e numpy
- 7. " Bringing Matplotlib to the Browser " ● Integra matplotlib com D3js ● Cria visualização de dados interativa no navegador mpld3
- 8. Vincent " A Python to Vega translator " " Vega is a visualization grammar, a declarative format for creating, saving and sharing visualization designs. " ● Estruturas de dados Python ● Visualização em JavaScript ● Integração com pandas
- 9. Bokeh ● Biblioteca de visualização interativa ● Integração com pandas ● Foco: explorar a visualização dos navegadores
- 10. dados: Top 100 Wines of the World 2014
- 12. Quantos vinhos por País? ggplot(wty, aes("Country")) + geom_bar(fill="red", colour="black") + theme(axis_text_x=element_text(angle=90))
- 13. Quantos vinhos por País? (com estilo) ggplot(wty, aes("Country")) + geom_bar(fill="red", colour="black") + theme_xkcd() + theme(axis_text_x=element_text(angle=90))
- 14. Qual a relação entre Pontos e Prêmios? ggplot(wty, aes(x='Prizes', y='Points')) + geom_point()
- 15. Qual a relação entre Pontos e Prêmios? ggplot(wty, aes(y="Points", x="Prizes")) + geom_point(size=60, shape=6) + stat_smooth(method='lm', colour='red')
- 16. Qual a relação entre Pontos e Prêmios por País? idx = (wty.Country == 'AUSTRALIA') | (wty.Country == 'FRANCE') | (wty.Country == 'SPAIN') | (wty.Country == 'USA') | (wty.Country == 'ARGENTINA') gplot(wty[idx], aes(x="Prizes", y="Points")) + geom_point() + stat_smooth(method='lm', color='blue') + facet_wrap("Country", scales="fixed")
- 17. Como se distribuem os prêmios por País? wty_country_prizes = wty['Prizes'].groupby(wty['Country']) idx = wty_country_prizes.count() == 1 wty_t = wty.ix[[c not in idx[idx].index for c in wty.Country], ['Country','Prizes']] wty_t_country_mean = wty_t.groupby(wty_t.Country).mean() country_names = wty_t_country_mean.Prizes.sort(inplace=False).index import seaborn as sb sb.set(style="ticks") f, ax = plt.subplots() sb.offset_spines() sb.violinplot(wty_t['Prizes'], wty_t['Country'], order=country_names) locs, labels = plt.xticks() plt.setp(labels, rotation=45) sb.despine(trim=True)
- 18. wty_country = wty.groupby('Country') country_count = wty_country['Wine'].aggregate(len) country_count.sort(inplace=True) fig, ax = plt.subplots(figsize=(12,8)) ax.spines['top'].set_visible(False) ax.spines['right'].set_visible(False) ax.spines['left'].set_visible(False) ax.get_xaxis().tick_bottom() ax.get_yaxis().tick_left() plt.subplot(221) g = wty_country['Prizes'].count() x = np.arange(len(g)) countries = list(g.index.values) y = g.values plt.bar(x+1, y, align='center', width=0.5, alpha=0.4) plt.xticks(x+1, [' ']*10, rotation=90) plt.title('Contagem') plt.ylabel('Prizes') plt.subplot(222) g = wty_country['Prizes'].mean() y = g.values plt.bar(x+1, y, align='center', color='red', width=0.5, alpha=0.4) plt.xticks(x+1, [' ']*10, rotation=90); plt.title(u'Média') plt.subplot(223) g = wty_country['Prizes'].max() y = g.values plt.bar(x+1, y, align='center', color='darkgreen', width=0.5, alpha=0.4) plt.xticks(x+1, countries, rotation=90); plt.title(u'Máx') plt.ylabel('Prizes') plt.subplot(224) g = wty_country['Prizes'].min() y = g.values plt.bar(x+1, y, align='center', color='magenta', width=0.5, alpha=0.4) plt.xticks(x+1, countries, rotation=90); plt.title(u'Mín') Mas qual País possui o melhor vinho?
- 19. >>> mpld3.save_html(fig, 'fig.html') <div id="fig_el672345700473765832459779"></div> <script> function mpld3_load_lib(url, callback){ var s = document.createElement('script'); s.src = url; s.async = true; s.onreadystatechange = s.onload = callback; s.onerror = function(){console.warn("failed to load library " + url);}; document.getElementsByTagName("head")[0].appendChild(s); } if(typeof(mpld3) !== "undefined" && mpld3._mpld3IsLoaded){ // already loaded: just create the figure !function(mpld3){ ... mpld3 salva fig em html
- 20. import Quandl tickers = [ "GOOG/BVMF_BOVA11.4", "YAHOO/SA_PETR3.6", "GOOG/BVMF_VALE5.4", "GOOG/BVMF_BBDC4.4", "GOOG/BVMF_BBAS3.4", "GOOG/BVMF_ITUB4.4"] series = Quandl.get(tickers, trim_start="2013-01-01", trim_end="2013-12-31") series.columns = [ "BOVESPA", "PETROBRAS", "VALE", "BRADESCO", "BB", "ITAU"] series.dropna(inplace=True) dados: Ações (Quandl) >>> print series BOVESPA PETROBRAS VALE BRADESCO BB ITAU Date 2013-01-02 61.27 19.00 42.60 36.02 25.80 31.3 2013-01-03 61.92 19.70 42.09 38.12 26.31 32.2 2013-01-04 61.32 19.77 41.17 37.45 26.00 31.9 2013-01-07 60.75 19.51 40.35 37.29 26.15 31.5 2013-01-08 59.91 18.96 40.10 37.42 26.45 31.6 2013-01-09 60.25 19.15 40.30 37.90 26.70 32.1 2013-01-10 60.49 19.27 40.57 37.40 26.30 31.7 2013-01-11 60.30 19.32 39.53 37.27 25.98 31.8 2013-01-14 60.75 19.21 39.77 37.88 26.93 32.2 2013-01-15 60.52 19.27 39.65 37.30 26.70 31.8 2013-01-16 60.54 19.34 39.68 37.49 26.86 32.1 2013-01-17 61.01 19.21 39.75 38.11 26.81 32.6 2013-01-18 60.79 19.05 39.20 38.25 26.20 32.7 2013-01-21 60.75 18.87 39.15 38.09 25.93 32.5 2013-01-22 60.55 19.12 39.59 38.40 25.64 32.9 ...
- 21. Visualização de Séries de Preço import vincent as v line = v.Line(series) line.axis_titles(x='Date', y='Price') line.legend(title='Series')
- 22. Exporta a Gramática de Gráficos >>> print line.to_json() { "axes": [ { "scale": "x", "title": "Date", "type": "x" }, { "scale": "y", "title": "Price", "type": "y" } ], "data": [ { "name": "table", "values": [ { "col": "BOVESPA", "idx": 1357092000000, "val": 61.27 }, { "col": "PETROBRAS","idx": 1357092000000, "val": 19.0 }, { "col": "VALE", "idx": 1357092000000, "val": 42.6 }, ...
- 23. Variação acumulada logreturns = np.log(series).diff().dropna(inplace=False) cumreturns = np.cumprod(np.exp(logreturns)) line = v.Line(cumreturns) line.axis_titles(x='Date', y='') line.legend(title='Series')
- 24. Séries de Preço Empilhadas stacked = v.StackedArea(series) stacked.axis_titles(x='Date', y='Price') stacked.legend(title='Series') stacked.colors(brew='Spectral')
- 25. Qual a correlação entre séries? # com seaborn sb.corrplot(np.log(series).diff().dropna())
- 26. Mais no detalhe sb.jointplot("VALE", "BOVESPA", series, kind="kde"); sb.jointplot("VALE", "BOVESPA", series, kind="reg");
- 27. Rentabilidade de ações ● Qual a rentabilidade de cada ação por mês? ● Como visualizar tudo no mesmo gráfico? ● Como fazer isso higienicamente? ● Vamos considerar 18 ações no ano de 2013
- 29. from bokeh.plotting import * output_notebook() import numpy as np import Quandl import pandas as pd from collections import OrderedDict tickers = ["GOOG/BVMF_BOVA11.4", "GOOG/BVMF_PETR3.4", "GOOG/BVMF_VALE5.4", 'GOOG/BVMF_BBDC4.4', 'GOOG/BVMF_BBAS3.4', 'GOOG/BVMF_ITUB4.4', 'GOOG/BVMF_USIM5.4', 'GOOG/BVMF_CIEL3. 4', 'GOOG/BVMF_CMIG4.4', 'GOOG/BVMF_CTIP3.4', 'GOOG/BVMF_CPLE6.4', 'GOOG/BVMF_ELET6.4', 'GOOG/BVMF_GGBR4.4', 'GOOG/BVMF_GOLL4.4', 'GOOG/BVMF_HGTX3.4', 'GOOG/BVMF_JBSS3.4', 'GOOG/BVMF_MRFG3.4', 'GOOG/BVMF_PDGR3.4'] series = Quandl.get(tickers, trim_start="2013-01-01", trim_end="2013-12-31", authtoken="nJ1NhTYdEs2p3MsS4CVd") series.columns = ['BOVESPA', 'PETROBRAS', 'VALE', 'BRADESCO', 'BB', 'ITAU', 'USIMINAS', 'CIELO', 'CEMIG', 'CETIP', 'COPEL', 'ELETROPAULO', 'GERDAU', 'GOL', 'HERING', 'JBS', 'MAFRIG', 'PDGR'] series.dropna(inplace=True) logreturns = np.log(series).diff().dropna() months = ["Jan", "Fev", "Mar", "Abr", "Mai", "Jun", "Jul", "Ago", "Set", "Out", "Nov", "Dez"] series_mr = logreturns.resample('M', how=sum)*100 series_mr['Month'] = [months[d.month-1] for d in series_mr.index] series_mr.set_index('Month', inplace=True) months = list(series_mr.index) stocks = list(series_mr.columns) import brewer2mpl bmap = brewer2mpl.get_map('RdYlGn', 'diverging', 3) colors = bmap.hex_colors levels = np.array([0, 1]) month = [] stock = [] color = [] rate = [] for y in stocks: for m in months: month.append(m) stock.append(y) monthly_rate = series_mr[y][m] rate.append(monthly_rate) color.append(colors[sum(levels < monthly_rate)]) source = ColumnDataSource(data=dict(month=month, year=stock, color=color, rate=rate, )) figure() rect('year', 'month', 0.95, 0.95, source=source, x_range=stocks, y_range=list(reversed(months)), color='color', line_color=None, tools="resize,hover,save", title="Rentabilidade de ações", plot_width=300, plot_height=400) grid().grid_line_color = None axis().axis_line_color = None axis().major_tick_line_color = None axis().major_label_text_font_size = "5pt" axis().major_label_standoff = 0 xaxis().location = "top" xaxis().major_label_orientation = np.pi/3 from bokeh.objects import HoverTool hover = [t for t in curplot().tools if isinstance(t, HoverTool)][0] hover.tooltips = OrderedDict([('date', '@month'), ('rate', '@rate'),]) show() Rentabilidade de ações Esse gráfico já não é tão simples! ;-)
- 30. Referências ● seaborn (github) ● vincent (github) ○ Trifacta - vega ● ggplot (github) ○ ggplot tutorial ○ ggplot2 no R ● AboutWilson.net: Refinando o estilo do matplotlib ● AboutWilson.net: Refinando o estilo do matplotlib com seaborn ● EuroPython 2014: Combining the powerful worlds of Python and R ● EuroPython 2014: Scientific Visualization with GR ● bokeh (github) ● mpld3 (github) ● prettyplotlib (github) ● plot.ly ● Quandl ● brewer2mpl (github)