pyspectra 0.0.1.2

A python package designed to work with spectroscopy data

Pyspectra

Welcome to pyspectra.
This package is intended to put functions together to analyze and transform spectral data from multiple spectroscopy instruments.

Currently supported input files are:

• .spc
• .dx

PySpectra is intended to facilitate working with spectroscopy files in python by using a friendly integration with pandas dataframe objects.
. Also pyspectra provides a set of routines to execute spectral pre-processing like:

• MSC
• SNV
• Detrend
• Savitzky - Golay
• Derivatives
• ..

Data spectra can be used for traditional chemometrics analysis but also can be used in general advanced analytics modelling in order to deliver additional information to manufacturing models by supplying spectral information.

#Import basic libraries
import spc
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn.decomposition import PCA

Read .spc file

Read a single file

from pyspectra.readers.read_spc import read_spc
spc=read_spc('pyspectra/sample_spectra/VIAVI/JDSU_Phar_Rotate_S06_1_20171009_1540.spc')
spc.plot()
plt.xlabel("nm")
plt.ylabel("Abs")
plt.grid(True)
print(spc.head())

gx-y(1)
908.100000    0.123968
914.294355    0.118613
920.488710    0.113342
926.683065    0.108641
932.877419    0.098678
dtype: float64

Read multiple .spc files from a directory

from pyspectra.readers.read_spc import read_spc_dir

df_spc, dict_spc=read_spc_dir('pyspectra/sample_spectra/VIAVI')
display(df_spc.transpose())
f, ax =plt.subplots(1, figsize=(18,8))
ax.plot(df_spc.transpose())
plt.xlabel("nm")
plt.ylabel("Abs")
ax.legend(labels= list(df_spc.transpose().columns))
plt.show()

gx-y(1)
gx-y(1)
gx-y(1)
gx-y(1)
gx-y(1)
gx-y(1)
gx-y(1)
gx-y(1)

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	JDSU_Phar_Rotate_S06_1_20171009_1540.spc	JDSU_Phar_Rotate_S11_2_20171009_1614.spc	JDSU_Phar_Rotate_S17_1_20171009_1652.spc	JDSU_Phar_Rotate_S23_1_20171009_1734.spc	JDSU_Phar_Rotate_S30_2_20171009_1815.spc	JDSU_Phar_Rotate_S37_2_20171009_1853.spc	JDSU_Phar_Rotate_S43_2_20171009_1928.spc	JDSU_Phar_Rotate_S49_1_20171009_2000.spc
908.100000	0.123968	0.164750	0.156647	0.147828	0.182833	0.171957	0.164471	0.149373
914.294355	0.118613	0.159980	0.150746	0.142974	0.178452	0.166827	0.159545	0.142818
920.488710	0.113342	0.155193	0.144959	0.138178	0.173734	0.161695	0.154330	0.136648
926.683065	0.108641	0.151398	0.140178	0.134014	0.170061	0.157110	0.149876	0.130452
932.877419	0.098678	0.141859	0.129715	0.124426	0.160590	0.147076	0.140119	0.119561
...	...	...	...	...	...	...	...	...
1651.422581	0.220935	0.262070	0.259643	0.242916	0.279041	0.271492	0.260664	0.252704
1657.616935	0.221848	0.262732	0.260664	0.243092	0.278962	0.272893	0.261647	0.254481
1663.811290	0.219904	0.260335	0.258975	0.240656	0.276382	0.271624	0.260278	0.253761
1670.005645	0.214080	0.253475	0.253110	0.234047	0.269528	0.265615	0.254568	0.248288
1676.200000	0.204217	0.242375	0.243082	0.223539	0.258771	0.255306	0.244826	0.238663

125 rows × 8 columns

Read .dx spectral files

Pyspectra is also built with a set of regex that allows to read the most common .dx file formats from different vendors like:

• FOSS
• Si-Ware Systems
• Spectral Engines
• Texas Instruments
• VIAVI

Read a single .dx file

.dx reader can read:

• Single files containing single spectra : read
• Single files containing multiple spectra : read
• Multiple files from directory : read_from_dir

Single file, single spectra

# Single file with single spectra
from pyspectra.readers.read_dx import read_dx
#Instantiate an object
Foss_single= read_dx()
# Run  read method
df=Foss_single.read(file='pyspectra/sample_spectra/DX multiple files/Example1.dx')
df.transpose().plot()

<matplotlib.axes._subplots.AxesSubplot at 0x1f44faa7940>

Single file, multiple spectra:

.dx reader stores all the information as attributes of the object on Samples. Each key represent a sample.

Foss_single= read_dx()
# Run  read method
df=Foss_single.read(file='pyspectra/sample_spectra/FOSS/FOSS.dx')
df.transpose().plot(legend=False)

<matplotlib.axes._subplots.AxesSubplot at 0x1f44f7f2e50>

for c in Foss_single.Samples['29179'].keys():
    print(c)

y
Conc
TITLE
JCAMP_DX
DATA TYPE
CLASS
DATE
DATA PROCESSING
XUNITS
YUNITS
XFACTOR
YFACTOR
FIRSTX
LASTX
MINY
MAXY
NPOINTS
FIRSTY
CONCENTRATIONS
XYDATA
X
Y

Spectra preprocessing

Pyspectra has a set of built in classes to perform spectra pre-processing like:

• MSC: Multiplicative scattering correction
• SNV: Standard normal variate
• Detrend
• n order derivative
• Savitzky golay smmothing

from pyspectra.transformers.spectral_correction import msc, detrend ,sav_gol,snv

MSC= msc()
MSC.fit(df)
df_msc=MSC.transform(df)


f, ax= plt.subplots(2,1,figsize=(14,8))
ax[0].plot(df.transpose())
ax[0].set_title("Raw spectra")

ax[1].plot(df_msc.transpose())
ax[1].set_title("MSC spectra")
plt.show()

SNV= snv()
df_snv=SNV.fit_transform(df)

Detr= detrend()
df_detrend=Detr.fit_transform(spc=df_snv,wave=np.array(df_snv.columns))

f, ax= plt.subplots(3,1,figsize=(18,8))
ax[0].plot(df.transpose())
ax[0].set_title("Raw spectra")

ax[1].plot(df_snv.transpose())
ax[1].set_title("SNV spectra")

ax[2].plot(df_detrend.transpose())
ax[2].set_title("SNV+ Detrend spectra")

plt.tight_layout()
plt.show()

Modelling of spectra

Decompose using PCA

pca=PCA()
pca.fit(df_msc)
plt.figure(figsize=(18,8))
plt.plot(range(1,len(pca.explained_variance_)+1),100*pca.explained_variance_.cumsum()/pca.explained_variance_.sum())
plt.grid(True)
plt.xlabel("Number of components")
plt.ylabel(" cumulative % of explained variance")

df_pca=pd.DataFrame(pca.transform(df_msc))
plt.figure(figsize=(18,8))
plt.plot(df_pca.loc[:,0:25].transpose())


plt.title("Transformed spectra PCA")
plt.ylabel("Response feature")
plt.xlabel("Principal component")
plt.grid(True)
plt.show()

Using automl libraries to deploy faster models

import tpot
from tpot import TPOTRegressor
from sklearn.model_selection import RepeatedKFold
cv = RepeatedKFold(n_splits=10, n_repeats=3, random_state=1)
model = TPOTRegressor(generations=10, population_size=50, scoring='neg_mean_absolute_error',
                      cv=cv, verbosity=2, random_state=1, n_jobs=-1)

y=Foss_single.Conc[:,0]
x=df_pca.loc[:,0:25]
model.fit(x,y)

HBox(children=(FloatProgress(value=0.0, description='Optimization Progress', max=550.0, style=ProgressStyle(de…



Generation 1 - Current best internal CV score: -0.30965836730187607

Generation 2 - Current best internal CV score: -0.30965836730187607

Generation 3 - Current best internal CV score: -0.30965836730187607

Generation 4 - Current best internal CV score: -0.308295313408046

Generation 5 - Current best internal CV score: -0.308295313408046

Generation 6 - Current best internal CV score: -0.308295313408046

Generation 7 - Current best internal CV score: -0.308295313408046

Generation 8 - Current best internal CV score: -0.3082953134080456

Generation 9 - Current best internal CV score: -0.3082953134080456

Generation 10 - Current best internal CV score: -0.3078569602146527

Best pipeline: LassoLarsCV(PCA(LinearSVR(input_matrix, C=0.1, dual=True, epsilon=0.1, loss=epsilon_insensitive, tol=0.01), iterated_power=3, svd_solver=randomized), normalize=False)





TPOTRegressor(cv=RepeatedKFold(n_repeats=3, n_splits=10, random_state=1),
              generations=10, n_jobs=-1, population_size=50, random_state=1,
              scoring='neg_mean_absolute_error', verbosity=2)

from sklearn.metrics import r2_score
r2=round(r2_score(y,model.predict(x)),2)
plt.scatter(y,model.predict(x),alpha=0.5, color='r')
plt.plot([y.min(),y.max()],[y.min(),y.max()],LineStyle='--',color='black')
plt.xlabel("y actual")
plt.ylabel("y predicted")
plt.title("Spectra model prediction R^2:"+ str(r2))

plt.show()