Scribd
Upload
5 views

Time series Assignment -1

The document discusses the components of a time series, including trend, seasonality, cyclical variations, and irregular variations, explaining their characteristics and types. It also highlights the importance of smoothing in time series analysis to identify trends and regular patterns, detailing methods such as moving average, exponential smoothing, and double exponential smoothing. Additionally, it compares forecasting methods like moving average, exponential smoothing, and linear regression, outlining their complexities, trend and seasonality detection capabilities, data requirements, and forecasting horizons.

Uploaded by

pramod walunj
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
5 views

Time series Assignment -1

The document discusses the components of a time series, including trend, seasonality, cyclical variations, and irregular variations, explaining their characteristics and types. It also highlights the importance of smoothing in time series analysis to identify trends and regular patterns, detailing methods such as moving average, exponential smoothing, and double exponential smoothing. Additionally, it compares forecasting methods like moving average, exponential smoothing, and linear regression, outlining their complexities, trend and seasonality detection capabilities, data requirements, and forecasting horizons.

Uploaded by

pramod walunj
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 3

PRN- 24079001186 SIU,Pune

Assignment 01: Time Series


Q1. Explain the components of a time series in detail:

 Trend
 Seasonality
 Cyclical Variations
 Irregular (Residual) Variations

1. Trends: It is referred to a long term upward or downward move- ment of the


data. In this case, the movement is considered to be monotonous for a
considerable period of time.
There are various types of trends which can be observed in a time series
data.
i. When a series of data shows a continuous increase over time,
then the trend is known as the upward trend.
ii. When a series of data shows a continuous decrease over time,
then the trend is known as the downward trend.
iii. When a series of data does not show any significant variation
over time, then the trend is known as the horizontal trend.
iv. When a series of data shows a complex or any random pattern
over time, then the trend is known as nonlinear trend.
2. Seasonality: It is referred to a specific pattern that gets repeated over a
particular interval. The interval can be a daily, weekly, monthly or yearly, here
are various types of trends which can be observed in a time series data.
a. A type of pattern repeated after every 7 days are known as Weekly
Seasonality example Data taken from sales, transporta- tion, etc.
follows such patterns.
b. A type of pattern repeated after every 30 or 31 days are known as
Monthly Seasonality example Data taken from energy bills, weather,
etc. follows such patterns.
c. A type of pattern repeated after every 365 or 366 days are known as
Yearly Seasonality eg. Data taken from agriculture, tourism, etc.
follows such patterns.
d. A type of pattern repeated after every particular Holiday or va-
cations are known as Holiday Seasonality eg. Data taken from
traffic, entertainment, etc. follows such patterns.
3. Cyclical Variation: Similar to Seasonality, Cyclic variation also shows the
regular patterns but are visible at an irregular period of time. These patterns
may occur due to various factors like economic cycles or any other
underlying patterns.
4. Irregular Variation: If the set of data follows an irregular pattern due to
irregular and unexpected fluctuations. These irregular pat- terns mainly
occur due to measurement errors or any type of noise causing sudden
fluctuations in the data.
Q2. Importance of Smoothing in Time Series Analysis.

Ans: Smoothing is a data pre-processing method used top avoid some


significant unexpected fluctuations from the set of data being varied over time.
This method helps to identify the trend or any type of regular patterns which
helps the statisticians to analyze the data. There are various types of smoothing
method mentioned below:
 Moving average smoothing: This method helps in removing the
fluctuations, by normalizing the present value as the averaged value of
the last k elements. This formula can be mathematically stated as:
 Exponential Smoothing: Unlike moving average in which the present
data of the time series were normalized by averaging the previous set of
data, Exponential Smoothing normalizes by assigning previous values
with exponentially decreasing weights. The smooth- ing is
mathematically represented as:
S0 = X0

St = α × Xt + (1 − α) × St−1 where t > 0, 0 < α < 1

 Double Exponential Smoothing: A method used to normalize the values


as well the trend of the data set is known as Double Expo- nential
smoothing. In this method, another parameter β is used to normalize the
trend of the dataset, whereas α is used to normalize the level of dataset.
Application:
To understand the application of Data Smoothing, let us consider an exam- ple
of Company accounting in which an Allowance for doubtful accounts are
prepared by using the data of bad debt expense from one reporting year to
another. If a Company is expecting to not receive the payment of certain goods
in two consecutive reporting years, example | 10000/- in first reporting period
and | 50000/- in second reporting period. In this case, the company may include
the total amount of | 60000/- if a high income has been expected in the first
reporting year. This assumption leads to rise in Income statement by | 60000/-
which will normalize the decrease in income by | 60000/-. This normalization
occurs because of data smoothing procedures.
Q3. Forecasting and Model Evaluation.
Ans. Compare and contrast the following forecasting methods:
Moving Average
Exponential Smoothing
Linear Regression
Parameters Moving Exponential Linear
Averages Smoothing Regression
Complexity Easy Moderate Moderate
Trend Poor detection Good detection Good but only
detection (due to Holt’s in case of Linear
Method) Trends
Seasonality Poor Good (due to Not used to
detection Holt-Winters detect
method) seasonality
Data Lagging Output is Output is
Sensitivity sensitive with sensitive with
respect to data respect to data
but if and only if
the data
follows linearity.
Data Stationary Data Stationary data, Stationary-linear
Requirements but may also work data.
well in data with
trends
or seasonality.
Horizons of Short Term Short-term to Medium-term to
Forecasting medium-term Long-term
Assumptions No assumptions Trends or Assumes linear
Seasonality relation between data
assumptions are and time.
made
Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root
Mean Square Error (RMSE) in assessing the performance of a time
series model.
• Mean Absolute Error is defined as the averaged magnitude
of the difference between actual and forecast value.
• Mean Squared Error is defined as the average squared
difference between the actual and forecast value. The
mathematical description

You might also like