機械学習によるPYS閾値予測のコード例

用意するもの

x:enegy　 y: PYS intensity or SPYS intensity

Excel(.xlsx) or CSV(.csv)

Data example(csv)

energy,pys
4.0,3.24
4.05,7.8
4.1,3.61
4.15,6.25
4.2,4.8
 .......

#うまく自作モジュールが読み込めないとき
import sys
sys.path.append('./pysfunclib')
sys.path.append('./autoreglib')

#　自作モジュールを変更したとき
%load_ext autoreload
%autoreload
%autoreload 2

#自作モジュールの読み込み
from pysfunclib import fowler_func as ff
from pysfunclib import fowler_func_opti as ffo
from pysfunclib import fit_prediction_lib as fpl
from pysfunclib import ml_prediction_lib as mpl
from pysfunclib import data_read as dr

#機械学習の自動化とプロットのモジュール
from autoreglib import gridreg as gs

# csv or excel dataの読み込み
#estimated value by analyst: 4.81eV
xx,yy= dr.read_pys_data("./data/df_au481_pys.csv",pys_or_spys='pys')

Input file path:  ./data/df_au481_pys.csv
Basename:  df_au481_pys.csv

MLによる予測

# 識別パラメータの入っているパスを指定してインスタンスを作成
prd=mpl.MLPredict(path_name='./spys_reg_20200623/')

# 識別パラメーターの読み込み
prd.param_load()

# データを与えて予測を行う
prd.prediction(xx,yy)

xl4070_005

{'gb': 5.06266670873689, 'rf': 5.143500000000013}

# データと予測結果の図示
prd.plot()

Fittingによる予測

fprd =  fpl.SPYSFit()

fprd.fit(xx, yy, para= [4.8,300,1,1])

{'abs_fit': array([ 0.90777292,  0.90777292,  0.90777292,  0.90777292,  0.90777292,
         0.90777292,  0.90777292,  0.90777292,  0.90777292,  0.90777292,
         0.90777292,  0.90777293,  0.90777296,  0.9077731 ,  0.90777366,
         0.90777608,  0.90778632,  0.90782978,  0.90801423,  0.90879656,
         0.91210747,  0.92599194,  0.98213669,  1.18304113,  1.72269947,
         2.73180203,  4.09262701,  5.64364671,  7.28641107,  8.97534068,
        10.68952135, 12.41869415, 14.15740673, 15.90253867, 17.65218798,
        19.40513314, 21.16055518, 22.91788496, 24.67671491, 26.43674562,
        28.19775224, 29.95956269, 31.72204302, 33.48508748, 35.2486114 ,
        37.01254625, 38.77683596, 40.54143419, 42.30630234, 44.07140799,
        45.83672369, 47.60222607, 49.36789512, 51.13371356, 52.89966643,
        54.6657407 , 56.43192498, 58.19820926, 59.96458468, 61.73104344,
        63.49757857]),
 'abs_para': array([  5.20536201, 401.39040875,   2.99107767,   0.82405168]),
 'abs_r2': 0.9866830175499015,
 'abs_ratio': 1.2477280140289777,
 'lsq_fit': array([ 1.56078152,  1.5847708 ,  1.61368739,  1.6484367 ,  1.69005051,
         1.73969193,  1.79865717,  1.86837437,  1.95039961,  2.04641144,
         2.15820525,  2.28768921,  2.43688349,  2.60792369,  2.80306883,
         3.02471353,  3.27540304,  3.55784936,  3.87494636,  4.22978145,
         4.62564136,  5.0660094 ,  5.55455156,  6.09508856,  6.6915512 ,
         7.34791627,  8.06812017,  8.85594496,  9.71486003, 10.64775676,
        11.65634554, 12.73938975, 13.88687785, 15.27257462, 16.52137294,
        17.90941608, 19.38811494, 20.94318237, 22.5682828 , 24.258699  ,
        26.00992277, 27.81742694, 29.6767045 , 31.58335221, 33.53314562,
        35.52209406, 37.54647574, 39.60285585, 41.68809109, 43.79932447,
        45.93397306, 48.08971144, 50.26445262, 52.45632796, 54.66366697,
        56.88497786, 59.11892913, 61.36433244, 63.62012704, 65.88536552,
        68.15920111]),
 'lsq_para': array([5.64999997e+00, 2.84935992e+03, 2.77124448e+02, 2.10181067e+00]),
 'lsq_r2': 0.9983609516424481,
 'lsq_ratio': 1.2557640312588878}

fprd.plot()

Loss function: MAE

---

Loss function: RMSE

---------END---------

# fprd.res_plot()