Time Sequence Information with NumPy

Picture by creativeart on Freepik

Time sequence knowledge is exclusive as a result of they rely on one another sequentially. It is because the information is collected over time in constant intervals, for instance, yearly, each day, and even hourly.

Time sequence knowledge are necessary in lots of analyses as a result of can symbolize patterns for enterprise questions like knowledge forecasting, anomaly detection, development evaluation, and extra.

In Python, you’ll be able to attempt to analyze the time sequence dataset with NumPy. NumPy is a strong package deal for numerical and statistical calculation, however it may be prolonged into time sequence knowledge.

How can we do this? Let’s attempt it out.
 

Time Sequence knowledge with NumPy

 
First, we have to set up NumPy in our Python atmosphere. You are able to do that with the next code should you haven’t achieved that.

Subsequent, let’s attempt to provoke time sequence knowledge with NumPy. As I’ve talked about, time sequence knowledge have sequential and temporal traits, so we might attempt to create them with NumPy.

import numpy as np

dates = np.array(['2023-01-01', '2023-01-02', '2023-01-03', '2023-01-04', '2023-01-05'], dtype="datetime64")
dates

Output>>
array(['2023-01-01', '2023-01-02', '2023-01-03', '2023-01-04',
       '2023-01-05'], dtype="datetime64[D]")

As you’ll be able to see within the code above, we set the information time sequence in NumPy with the dtype parameter. With out them, the information can be thought of string knowledge, however now it’s thought of time sequence knowledge.

We are able to create the NumPy time sequence knowledge with out writing them individually. We are able to do this utilizing the sure methodology from NumPy.

date_range = np.arange('2023-01-01', '2025-01-01', dtype="datetime64[M]")
date_range

Output>>
array(['2023-01', '2023-02', '2023-03', '2023-04', '2023-05', '2023-06',
       '2023-07', '2023-08', '2023-09', '2023-10', '2023-11', '2023-12',
       '2024-01', '2024-02', '2024-03', '2024-04', '2024-05', '2024-06',
       '2024-07', '2024-08', '2024-09', '2024-10', '2024-11', '2024-12'],
      dtype="datetime64[M]")

We create month-to-month knowledge from 2023 to 2024, with every month’s knowledge because the values.

After that, we will attempt to analyze the information primarily based on the NumPy datetime sequence. For instance, we will create random knowledge with as a lot as our date vary.

knowledge = np.random.randn(len(date_range)) * 10 + 100 

Output>>
array([128.85379394,  92.17272879,  81.73341807,  97.68879621,
       116.26500413,  89.83992529,  93.74247891, 115.50965063,
        88.05478692, 106.24013365,  92.84193254,  96.70640287,
        93.67819695, 106.1624716 ,  97.64298602, 115.69882628,
       110.88460629,  97.10538592,  98.57359395, 122.08098289,
       104.55571757, 100.74572336,  98.02508889, 106.47247489])

Utilizing the random methodology in NumPy, we will generate random values to simulate time sequence evaluation.

For instance, we will attempt to carry out a shifting common evaluation with NumPy utilizing the next code.

def moving_average(knowledge, window):
    return np.convolve(knowledge, np.ones(window), 'legitimate') / window

ma_12 = moving_average(knowledge, 12)
ma_12

Output>>
array([ 99.97075433,  97.03945458,  98.20526648,  99.53106381,
       101.03189965, 100.58353316, 101.18898821, 101.59158114,
       102.13919216, 103.51426971, 103.05640219, 103.48833188,
       104.30217122])

Shifting common is a straightforward time sequence evaluation by which we calculate the imply of the subset variety of the sequence. Within the instance above, we use window 12 because the subset. This implies we take the primary 12 of the sequence because the subset and take their means. Then, the subset strikes by one, and we take the subsequent imply subset.

So, the primary subset is that this subset the place we takes the imply:

[128.85379394,  92.17272879,  81.73341807,  97.68879621,
       116.26500413,  89.83992529,  93.74247891, 115.50965063,
        88.05478692, 106.24013365,  92.84193254,  96.70640287]

The following subset is the place we slide the window by one:

[92.17272879,  81.73341807,  97.68879621,
       116.26500413,  89.83992529,  93.74247891, 115.50965063,
        88.05478692, 106.24013365,  92.84193254,  96.70640287,
        93.67819695]

That’s what the np.convolve does as the strategy would transfer and sum the sequence subset as a lot because the np.ones array quantity. We use the legitimate possibility solely to return the quantity that may be calculated with none padding.

However, shifting averages are sometimes used to investigate time sequence knowledge to determine the underlying sample and as alerts comparable to purchase/promote within the monetary discipline.

Talking of patterns, we will simulate the development knowledge in time sequence with NumPy. The development is a long-term and chronic directional motion within the knowledge. Mainly, it’s the basic course of the place the time sequence knowledge can be.

development = np.polyfit(np.arange(len(knowledge)), knowledge, 1)
development

Output>>
array([ 0.20421765, 99.78795983])

What occurs above is we match a linear straight line to our knowledge above. From the end result, we get the slope of the road (first quantity) and the intercept (second quantity). The slope represents how a lot knowledge adjustments per step or temporal values on common, whereas the intercept is the information course (constructive is upward and unfavourable is downward).

We are able to even have detrended knowledge, that are the parts after we take away the development from the time sequence. This knowledge kind is commonly used to detect fluctuation patterns within the development knowledge and anomalies.

detrended = knowledge - (development[0] * np.arange(len(knowledge)) + development[1])
detrended

Output>>
array([ 29.06583411,  -7.81944869, -18.46297706,  -2.71181657,
        15.66017371, -10.96912278,  -7.2707868 ,  14.29216727,
       -13.36691409,   4.61421499,  -8.98820376,  -5.32795108,
        -8.56037465,   3.71968235,  -5.00402087,  12.84760174,
         7.8291641 ,  -6.15427392,  -4.89028352,  18.41288776,
         0.6834048 ,  -3.33080706,  -6.25565918,   1.98750918])

The info with out their development are proven within the output above. In a real-world software, we might analyze them to see which one deviates an excessive amount of from the widespread sample.

We are able to additionally attempt to analyze seasonality from the time sequence knowledge we’ve. Seasonality is the common and predictable patterns that happen at particular temporal intervals, comparable to each 3 months, each 6 months, and others. Seasonality is often affected by exterior components comparable to holidays, climate, occasions, and lots of others.

seasonality = np.imply(knowledge.reshape(-1, 12), axis=0)
seasonal_component = np.tile(seasonality, len(knowledge)//12 + 1)[:len(data)]

Output>>
array([111.26599544,  99.16760019,  89.68820205, 106.69381124,
       113.57480521,  93.4726556 ,  96.15803643, 118.79531676,
        96.30525224, 103.4929285 ,  95.43351072, 101.58943888,
       111.26599544,  99.16760019,  89.68820205, 106.69381124,
       113.57480521,  93.4726556 ,  96.15803643, 118.79531676,
        96.30525224, 103.4929285 ,  95.43351072, 101.58943888])

Within the code above, we calculate the typical for every month after which lengthen the information to match its size. In the long run, we get the typical for every month within the two-year interval, and we will attempt to analyze the information to see if there may be seasonality price mentioning.

That’s all the essential methodology we will do with NumPy for time sequence knowledge and evaluation. There are a lot of superior strategies, however the above is the essential we will do.
 

Conclusion

 
The time sequence knowledge is a singular knowledge set because it represents in a sequential method and has temporal properties. Utilizing NumPy, we will set the time sequence knowledge whereas performing primary time sequence evaluation comparable to shifting averages, development evaluation, and seasonality evaluation. knowledge whereas performing primary time sequence evaluation comparable to shifting averages, development evaluation, and seasonality evaluation.
 
 

Cornellius Yudha Wijaya is an information science assistant supervisor and knowledge author. Whereas working full-time at Allianz Indonesia, he likes to share Python and knowledge suggestions through social media and writing media. Cornellius writes on quite a lot of AI and machine studying matters.