Webinar on Forecasting Techniques

Webinar: Forecasting and Sampling Techniques

• Webinar title: SAKSHAM: IL6 Webinar for Logistics, "Forecasting and Sampling Techniques"
• Presenter: Priyendu N Giri
• Contact: 7077762551
• Company: Tata Steel

Basics of Statistics

• Types of Data:
  • Numerical (Discrete & Continuous)
  • Categorical (Nominal & Ordinal)
• Statistical Measures:
  • Mean, Median, Mode
  • Variance, Standard Deviation
  • Dependent and Independent Variables
  • Correlation

Decisions Requiring Forecasting

• Examples:
  • Which markets to pursue?
  • What products to produce?
  • How many people to hire?
  • How many units to purchase?
  • How many units to produce?
  • ...and more...

Common Characteristics of Forecasting

• Forecasts are rarely perfect
• Forecasts are more accurate for aggregated data than for individual items
• Forecasts are more accurate for shorter than longer time periods

Forecasting Steps

• Define what needs to be forecast (level of detail, units of analysis, time horizon)
• Identify available data and evaluate its suitability
• Select and test a forecasting model (considering cost, ease of use, and accuracy)
• Generate the forecast
• Monitor forecast accuracy over time

Types of Forecasting Models

• Qualitative: Forecasts generated subjectively by the forecaster
• Quantitative (Statistical): Forecasts generated through mathematical modeling

Qualitative Forecasting Models

• Executive opinion: A group of managers provides a forecast
• Market research: Surveys and interviews used to identify customer preferences
• Delphi method: Eliciting consensus among experts

Statistical Forecasting Models

• Time Series Models:
  • Assume future follows past patterns
  • Single variable models
  • Machine learning models
• Causal Models:
  • Explore cause-and-effect relationships
  • Use leading indicators to predict the future
  • Use independent/feature variables to forecast dependent/target variables

Composition of Time Series Data

• Data = historic pattern + random variation
• Historic pattern may include:
  • Level (long-term average)
  • Trend
  • Seasonality
  • Cycle

Time Series Patterns

• Level or Horizontal Pattern: Data varies around a consistent mean value
• Trend Pattern: Data shows a consistent upward or downward movement over time
• Seasonal Pattern: Data exhibits a regularly repeating pattern over time
• Cycle Pattern: Data displays a repetitive but non-fixed pattern of increases and decreases over a longer period

Level Forecasting Methods

• Naïve Forecasting
• Simple Mean
• Moving Average
• Weighted Moving Average
• Exponential Smoothing

Time Series Example Problem

• Includes data and demonstrates various forecasting methods

Forecast Accuracy

• Forecasts are rarely perfect
• Need to measure forecast error (E = A - F)
• Over forecasts: negative error
• Under forecasts: positive error

Tracking Forecast Error

• Mean Absolute Deviation (MAD)
• Mean Square Error (MSE)

Accuracy and Tracking Signal Example:

• Data comparison, Methods A & B
• Measurement of MAD and MSE
• Tracking Signal
• Review of forecast accuracy

Sampling Techniques

• Population: Collection of elements we wish to make inferences about
• Sampling Units: Non-overlapping collections of elements that cover the entire population
• Probability Sampling:
  • Each member has a known non-zero probability of selection
  • Random sampling, systematic sampling, stratified sampling
• Non-probability Sampling:
  • Selection is not random; includes convenience sampling, judgment sampling, quota sampling, snowball sampling

Sampling for Population

• Study Population: Collection of subjects relevant to the investigation
• Target Population: The wider group being studied, which data generalizes to
• Sampling Frame: A complete list of all elements in the study population
• Sample: A subset of the study population selected for analysis

Simple Random Sampling

• Purest form of probability sampling
• Each member has an equal chance of selection
• Useful when populations are manageable

Systematic Random Sampling

• Every nth member is selected
• Good when a complete list is available
• May not be truly random if underlying order exists

Stratified Random Sampling

• Population segmented by strata (groups)
• Proportional representation ensures all strata are adequately represented
• Varied sampling fractions within strata are possible; important in minority groups

Cluster Sampling

• Population divided into clusters
• Random sample of clusters is selected
• All or some units are then selected from the selected clusters
• Useful when a complete list is impractical

Multistage Sampling

• Complex cluster sampling; multiple stages of random sampling
• Subsamples from clusters are randomly selected at each stage
• Useful for larger populations where complete list is missing

Multiphase Sampling

• Collecting information from whole & part of a sample at different phases
• Useful for costly or complex data collection processes

Convenience Sampling

• Selection based on ease and accessibility

Judgmental or Purposive Sampling

• Selection based on researcher judgment
• Researcher must ensure sample reflects population

Quota Sampling

• Aims for proportion in sub-groups; like stratified; but selection non-random

Snowball Sampling

• Initial subjects recruit additional subjects
• Useful for rare or hidden populations