DATA4400 Assessment 3 Quiz

Section 1: Data Selection for Time-Series Business Problems

Question 1: E-commerce Revenue Analysis

Context: An online retailer wants to forecast quarterly revenue for the next 2 years to plan inventory and marketing budgets.

Based on the revenue data shown above, what is the most appropriate time-series characteristic to consider for forecasting?

a) The data shows no clear pattern, so use a simple moving average b) Strong seasonal pattern with Q4 peaks suggests seasonal ARIMA or Holt-Winters c) Linear trend only, use simple linear regression d) Random walk pattern, use exponential smoothing

Question 2: Restaurant Customer Traffic

Day of Week	Avg Daily Customers	Variance	Trend (6 months)
Monday	120	25	+5%
Tuesday	110	20	+3%
Wednesday	115	22	+4%
Thursday	125	28	+6%
Friday	180	35	+8%
Saturday	220	45	+10%
Sunday	200	40	+9%

A restaurant chain wants to optimize staff scheduling. What frequency of data collection would be most appropriate for this forecasting problem?

a) Monthly aggregated data to smooth out daily variations b) Daily data to capture weekly seasonality and day-of-week effects c) Hourly data to capture peak dining times d) Quarterly data to identify long-term trends only

Section 2: Extracting Business Insights from Time-Series Data

Question 3: Energy Consumption Pattern Analysis

Context: A manufacturing company's monthly energy consumption over 3 years.

What key business insight can be extracted from this energy consumption pattern?

a) Energy usage is random and unpredictable b) Clear seasonal pattern with winter peaks suggests heating costs drive consumption c) Declining trend indicates equipment efficiency improvements d) Stable consumption suggests no operational changes needed

Question 4: Social Media Engagement Metrics

Month	Posts	Likes	Shares	Comments	Engagement Rate
Jan 2024	45	1,250	89	156	3.2%
Feb 2024	52	1,680	134	203	4.1%
Mar 2024	48	1,920	167	245	4.8%
Apr 2024	43	1,456	98	178	3.9%
May 2024	51	2,150	201	287	5.2%
Jun 2024	46	1,834	143	234	4.6%

Based on the social media data above, what is the most actionable business insight for content strategy?

a) More posts always lead to higher engagement b) Engagement rate is increasing despite posting frequency variations, suggesting content quality improvements c) Comments are the least important metric to track d) The data shows no clear patterns for strategy decisions

Section 3: Understanding Business Problems in Forecasting Context

Question 5: Supply Chain Optimization

Scenario: A retail chain experiences stockouts during peak seasons and excess inventory during low seasons. The chart shows monthly inventory levels vs. demand.

What is the core business problem that time-series forecasting should address?

a) Reduce overall inventory levels to cut costs b) Align inventory procurement timing with seasonal demand patterns to minimize stockouts and overstock c) Increase inventory across all months to avoid stockouts d) Focus only on demand forecasting without considering inventory timing

Question 6: Healthcare Resource Planning

Department	Current Capacity	Peak Demand	Average Utilization	Seasonal Pattern
Emergency	50 beds	62 patients	78%	Winter flu season
Surgery	20 rooms	24 surgeries	85%	Summer elective surge
Cardiology	30 beds	35 patients	82%	Post-holiday stress
Maternity	25 beds	22 births	65%	Spring baby boom

A hospital needs to optimize staff allocation across departments. What is the primary business problem for forecasting?

a) Reduce staff costs by cutting personnel across all departments b) Predict department-specific demand patterns to enable dynamic staff scheduling and capacity planning c) Maintain constant staffing levels regardless of demand variations d) Focus only on the emergency department as it has highest utilization

Section 4: Applying Time-Series Forecasting Methods

Question 7: Tourism Demand Forecasting

Context: Monthly tourist arrivals showing strong seasonality and growing trend. Data exhibits both multiplicative seasonality and irregular events (e.g., COVID-19 impact).

Which forecasting method would be most appropriate for this tourism data?

b) Simple exponential smoothing (no trend/seasonality) b) Holt-Winters multiplicative method to handle trend and multiplicative seasonality c) Moving average to smooth out fluctuations d) Linear regression with time as the only predictor

Question 8: Financial Stock Price Analysis

Model Type	Data Requirements	Best For	Assumptions
ARIMA(p,d,q)	Stationary series	Non-seasonal trends	Linear relationships
SARIMA	Seasonal patterns	Seasonal time series	Constant seasonality
VAR	Multiple variables	Multivariate analysis	All variables stationary
Prophet	Strong seasonality	Business applications	Handles missing data

A financial analyst needs to forecast daily stock returns that show volatility clustering and no clear seasonal pattern. Which model approach is most suitable?

a) SARIMA for seasonal patterns b) ARIMA after testing for stationarity and determining appropriate (p,d,q) parameters c) Prophet for business seasonality d) VAR model for single variable analysis

Section 5: Evaluating Forecasting Models

Question 9: Model Performance Comparison

Model	MAPE (%)	RMSE	MAE	AIC	Training Time
Simple Exponential Smoothing	12.5	145.2	98.7	1,234	2 sec
Holt-Winters	8.3	102.4	76.3	1,156	15 sec
ARIMA(2,1,2)	9.1	118.7	82.1	1,189	45 sec
Prophet	7.8	98.5	71.2	1,142	30 sec
Seasonal ARIMA	8.9	105.3	78.9	1,167	120 sec

Based on the model evaluation metrics above, which model performs best overall for a business forecasting application?

a) Simple Exponential Smoothing (fastest training time) b) Prophet (best balance of accuracy metrics with reasonable complexity) c) Seasonal ARIMA (most sophisticated model) d) Holt-Winters (lowest AIC value)

Question 10: Cross-Validation Strategy

Context: Time-series cross-validation showing rolling window approach with multiple train/test splits.

Why is time-series cross-validation different from standard k-fold cross-validation?

a) No difference - both methods are equivalent b) Time-series CV maintains temporal order and uses forward-chaining to avoid data leakage from future to past c) Time-series CV uses fewer folds to save computation time d) Time-series CV randomly shuffles data for better generalization

Section 6: Data-Driven Business Impact Recommendations

Question 11: Retail Inventory Optimization Impact

Scenario	Current Situation	With Forecasting	Potential Impact
Stockout Rate	15%	5%	+$2.3M revenue
Excess Inventory	$5.2M	$2.1M	+$3.1M cash flow
Storage Costs	$850K/month	$620K/month	+$2.76M/year
Staff Overtime	2,400 hrs/month	800 hrs/month	+$960K/year

Based on this forecasting impact analysis, what is the most compelling business case for implementing time-series forecasting?

a) Focus only on reducing stockout rates for customer satisfaction b) Combined impact of $9M+ annual improvement through better demand prediction and inventory optimization c) Implement forecasting primarily to reduce storage costs d) Use forecasting only for staff scheduling optimization

Question 12: Marketing Budget Allocation

Context: Marketing ROI by channel over 18 months, showing seasonal patterns and effectiveness trends.

What data-driven recommendation would you make for next quarter's marketing budget allocation?

a) Distribute budget equally across all channels b) Increase digital and social media allocation based on improving ROI trends, reduce traditional media during low-season periods c) Focus all budget on the single highest-performing channel d) Maintain current allocation without considering seasonal patterns

Time Series Forecasting & Business Applications