a
L9›fç% ã @ s€ d dl Zd dlZd dlmZ d dlmZ d dl Z
d dl Zd dl m Z m
Z
d dl Zd dlmZ d dl mZ dZG dd„ deƒZdS ) é N)Ú HydraHeadApp)ÚStepsÚStep)Úchain)ÚSequencyz{:,.2f}c @ s e Zd Zddd„Zdd„ ZdS )ÚWalshAppÚ c K s | j |¡ || _d S )N)Ú__dict__ÚupdateÚtitle)Úselfr Úkwargs© r úF/Users/putuwistika/Documents/TRILAK/hydralit-example/apps/walsh_app.pyÚ__init__ s zWalshApp.__init__c
C sÎ zxt d¡ t d¡ t jddd tjdddgdd }| |d
d ¡ t d ¡ t d
¡ t j j
dddddd}t j j
dddddd}t j dd¡}t ƒ }t
j ddd¡}tƒ jttdt|ƒƒƒ|d}|j| ¡ ||d }|j| ¡ ||d } tƒ j| ¡ |d}
tƒ j| ¡ | d} |jd d } | j| d d d!d" |
j| d#d d$d%d& | ¡ t j| ¡ dd' t d(¡ t d)¡ tj |ddd*}
|
j t d+d,d¡d-}|j| ¡ ||d }|j| ¡ ||d }tƒ j|jdd.|d}tƒ j|jdd.|d}|jd d/d0} |j| d1d2d3 |j| d1d4d3 | ¡ t j| ¡ dd' t d5¡ t d6¡ t j d,d7d8\}\} }}|j!d9d: t ƒ }| "| ¡ ¡\}}}| #d;¡ | $d<¡ | %d=¡ | &||¡ | ¡ j'd> }| (| ¡ d t
j) | ¡\}} | #d?¡ | $d@¡ | %d=¡ | &|| ¡ | (| *¡ d t
j) | ¡\} } | #dA¡ | $d@¡ | %d=¡ | &| dd … | dd … ¡ t j|dd' W nN t+yÈ } z4t j,dBddC t -dD¡ t -dE .| ¡¡ W Y d } ~ n
d } ~ 0 0 d S )FNzSequency Denoisinga‡ Source for this quick tutorial on using Sequency methods for denoising step change data can be found in the [Hotstepper documentation](https://hotstepper.readthedocs.io/notebooks/sequency_quickstart.html). An example of how easy it is to convert an existing application and use within a Hydralit multi-page application, see the secret saurce [here] (https://github.com/TangleSpace/hydralit).z
T)Úunsafe_allow_htmlzZhttps://raw.githubusercontent.com/TangleSpace/hotstepper-data/master/data/vessel_queue.csvZenterZleave)Ú parse_datesÚdayfirstzdata.csvzDownload Example Dataa<
Another typical use case of Fourier transforms is to remove high frequency noise from a single by decomposing it into constituent frequency components and setting frequencies below a threshold value to zero. With Sequency analysis, we have a similar functionality, except since we have step data and we are wanting to retain the nature of the step data (as steps) instead of just smoothing away the details, we can use the denoise method of the Sequency module to remove the higher sequency number components. An explicit example is shown below, here for now, we pass in the direct steps data for denoising, apply a strength parameter to determine how many of the high sequency components are set to zero and then we construct a new Steps object.
a¥
denoise_step_changes_strong = seq.denoise(rand_steps.step_changes(),denoise_strength=0.7,denoise_mode='range')
denoise_step_changes = seq.denoise(rand_steps.step_changes(),denoise_strength=0.2)
rand_steps_denoise_strong = Steps().add_direct(data_start=rand_steps.step_keys(),data_weight = denoise_step_changes_strong)
rand_steps_denoise = Steps().add_direct(data_start=rand_steps.step_keys(),data_weight = denoise_step_changes)
ax = rand_steps.plot(label='Original Data')
rand_steps_denoise.plot(ax=ax,color='g',linewidth=2,label='Light Denoise');
rand_steps_denoise_strong.plot(ax=ax,color='r',linewidth=2,linestyle='-',label='Strong Denoise')
ax.legend();
zStrong Denoise Strengthgffffffæ?g{®Gáz„?g ð?)ÚvalueÚ min_valueÚ max_valueÚstepzLight Denoise Strengthgš™™™™™É?z Denoise Mode)Úranger iöÿÿÿé éd é )Z
data_startZ data_weight)Zdenoise_strengthZ denoise_modez
Original Data)ÚlabelÚgé z
Light Denoise)ÚaxÚcolorÚ linewidthr ÚrÚ-zStrong Denoise)r r r! Z linestyler )Ú clear_figureuO
As another quick example, we can apply the same technique to one of the HotStepper sample datasets, for this tutorial, we’ll only look at the first two months of data in order to better highlight what the sequency denoising can do to help better show the trend or typical step behaviour of the data.
aQ
df_vq_samp = pd.read_csv(r'https://raw.githubusercontent.com/TangleSpace/hotstepper-data/master/data/vessel_queue.csv',parse_dates=['enter','leave'],dayfirst=True)
vq_samp = Steps.read_dataframe(df_vq_samp,start='enter',end='leave')
vq_clip = vq_samp.clip(ubound=pd.Timestamp(2020,3,1))
dn = seq.denoise(vq_clip.step_changes(),denoise_mode='range')
vq_clip_dn = Steps().add_direct(data_start=vq_clip.step_keys(convert_keys=True),data_weight=dn)
ax = vq_clip.plot(label='Original Data',figsize=(14,6))
vq_clip_dn.plot(ax=ax,linewidth=3,label='Sequency Denoise')
ax.legend()
)ÚstartÚendiä é )Zubound)Z convert_keys)é é )r Úfigsizeg ø?zStrong Sequency Denoise)r r! r zLight Sequency Denoisea’
As the last item, we can take a look at the sequency spectrum for the vessel queue data. This dataset has a large number of changes and therefore the sequency range goes quite high, however it does show a number of peaks that are significantly larger than the others, indicating a number of distinct and repeating step change patterns within the data over this period.
aÌ
fig, (ax,ax2,ax3) = plt.subplots(nrows=3,figsize=(12,8))
fig.tight_layout(h_pad=4)
# Sequency object to use for analysis
vq_clip_seq = Sequency()
n,sp,l = vq_clip_seq.sequency_spectrum(vq_clip.step_changes())
ax.set_title('Step Change Sequency Spectrum')
ax.set_xlabel('Sequency Number')
ax.set_ylabel('Amplitude')
ax.stem(n,sp)
# number of data points, needed to create the sampling frequency
no_points = vq_clip.step_changes().shape[0]
fr,fsp = vq_clip_seq.frequency_spectrum(vq_clip.step_changes(),2*np.pi*no_points)
ax2.set_title('Step Change Frequency Spectrum')
ax2.set_xlabel('Frequency')
ax2.set_ylabel('Amplitude')
ax2.stem(fr,fsp)
# FFT the steps values instead of the delta changes to see the difference in the spectrum.
frv,fspv = vq_clip_seq.frequency_spectrum(vq_clip.step_values(),2*np.pi*no_points)
ax3.set_title('Steps Value Frequency Spectrum')
ax3.set_xlabel('Frequency')
ax3.set_ylabel('Amplitude')
ax3.stem(frv[1:],fspv[1:]);
)r é )Únrowsr* é )Zh_padz Step Change Sequency SpectrumzSequency NumberZ Amplituder z Step Change Frequency SpectrumÚ Frequencyz Steps Value Frequency Spectrumz./resources/failure.png)ÚwidthzhAn error has occurred, someone will be punished for your inconvenience, we humbly request you try again.zError details: {})/Ústr Ú subheaderÚmarkdownÚpdÚread_csvÚdownload_buttonÚwriteÚcodeÚsidebarÚsliderÚradior ÚnpÚrandomÚrandintr Z
add_directÚlistr ÚlenZdenoiseZ step_changesZ step_keysÚplotÚlegendÚpyplotZ
get_figureZread_dataframeÚclipÚ TimestampÚpltÚsubplotsZ tight_layoutZsequency_spectrumZ set_titleZ
set_xlabelZ
set_ylabelÚstemÚshapeZfrequency_spectrumÚpiZ step_valuesÚ ExceptionÚimageÚerrorÚformat) r Z
df_vq_sampZdenoise_power_strong_selectZdenoise_power_light_selectZdenoise_mode_selectÚseqZ
steps_changesZ
rand_stepsZdenoise_step_changes_strongZdenoise_step_changesZrand_steps_denoise_strongZrand_steps_denoiser Zvq_sampZvq_clipZ dn_strongZdn_lightZvq_clip_dn_strongZvq_clip_dn_lightÚfigÚax2Zax3Z vq_clip_seqÚnÚspÚlZ no_pointsÚfrZfspZfrvZfspvÚer r r Úrun s”
ÿ ÿ ÿÿÿÿ$
z WalshApp.runN)r )Ú__name__Ú
__module__Ú __qualname__r rV r r r r r s
r )Úpandasr3 Ú streamlitr0 Úhydralitr Úmatplotlib.pyplotrB rE Únumpyr; Z
hotstepperr r ZhsÚ itertoolsr r ZSIG_FIGURE_FMTr r r r r Ú s