× Limited Time Offer ! FLAT 20-40% off - Grab Deal Before It’s Gone. Order Now
Connect With Us
Order Now

Data4400 Data Driven Decision Making and Forecasting IT Report Sample

Your Task

Apply forecasting techniques to a given dataset and provide a business application of the forecasts. The report is worth 30 marks (see rubric for allocation of these marks).

Assessment Description

A dataset from a retailer that has more than 45 stores in different regions (Public data from Kaggle) has been sourced. The data provided for the assessment represents two stores. Store number 20 has the highest revenue within the country and store 25 does not have a high volume of sales. The objective of the assessment is to develop different demand forecast models for these stores and compare the forecast models in terms of accuracy, trend, and seasonality alignment with the historical data provided. Students must use visual inspection, error metrics and information criteria on the test data to provide conclusions.

Assessment Instructions

In class: You will be presented with a dataset in class. As a group, analyse the dataset using Tableau and Exploratory.io. You will provide an oral presentation of the group work in parts A to C during the third hour of the workshop.

The data set will be posted or emailed to you at the beginning of class in Week 6.

After Class: Individually write a 1000-word report which briefly summarises the analysis and provides suggestions for further analysis. This component of the assessment is to be submitted via Turnitin in by Tuesday of week 7. No marks will be awarded for the assessment unless this report is submitted.

Hint: take notes during the group assessment to use as prompts for your report.As a group:

Part A

- Use Tableau to compare the two stores in terms of sales using adequate visualisation(s).
- Run Holt-Winters forecasts of the next 5 months for stores 20 and 25.
analyse the results of the forecasts in terms of:
o Accuracy
o Alignment with the historical trend
o Alignment with the historical seasonality

Part B

- Use Exploratory to generate ARIMA forecasts for stores 20 and 25.
- Create visualisations, interpret and describe your findings.
- Analyse the forecasts in terms of:
o Accuracy
o Alignment with the historical trend.
o Alignment with the historical seasonality.

Part C

Prepare a presentation:
• Include key findings.
• Highlight methodologies.
• Advise which methods to use for each store.
• Recommend improvements in terms of forecasting for the retailer.

Note: All members of the group should be involved in the presentation. The allocated time for the presentation will be decided by your lecturer.

Solution

Introduction

The ability for organisations to base decisions on empirical evidence rather than preconceptions makes data-driven decision-making and forecasting essential. For Assignment Help, Forecasting trends help proactive tactics, resource optimisation, and market leadership in fast-moving environments. With the aid of various forecasting models, including ARIMA, HOLT-WINTERS, and others, the study's goal is to visualise the sales of both STORE 20 and STORE 25 and forecast sales based on historical sales trends.
Discussion on Results

Figure 1: Visualization of STORE 25 sales

Figure 2: Forecast result of STORE 25 sales

With a decline of 336,738 units from the initial figure of 3,149,931 units in October 2012, the projection for STORE 25 sales from October 2012 to February 2013 shows a downward trend. With a peak in December 2012 (1,616,475) and a trough in January 2013 (-563,853 units), the seasonal effect is clear.

Figure 3: Sum of value for STORE 25 sales

It appears reasonable to use the selected additive model for level, trend, and seasonality. The forecast's accuracy is fairly high, with a low MAPE of 10.8%, despite occasional forecast errors, as seen by measures like RMSE (383,833) and MAE (296,225). This shows that the model effectively captures underlying patterns, assisting in the formulation of successful decisions for the STORE 25 sales strategy.

Figure 4: Visualization of STORE 20 sales

Figure 5: Forecast result of STORE 20 sales

A time series methodology was used to determine the sales prediction for STORE 20 for the period from October 2012 to February 2013. Notably, it was clear that an additive model for level and trend had been used and that there was no identifiable seasonal regularity. Sales began at roughly $9.88 million in October 2012, and by February 2013, they had increased by $197,857.

Figure 6: Sum of value for STORE 20 sales

Quality metrics showed an RMSE of $1.3 million and a fair degree of accuracy. The forecast's relative accuracy may be seen in the forecast's mean absolute percentage error (MAPE), which was 12.4%. STORE 20's sales trend could be understood by the chosen model despite the lack of a pronounced seasonal effect.

Figure 7: Visualization of HOLT-WINTERS test for STORE 25 sales

Figure 8: Result of HOLT-WINTERS test for STORE 25 sales

When five periods of STORE 25 sales data are smoothed using the HOLT-WINTERS exponential method, a downward trend is evident. The anticipated values start at 3,028,050.52 and successively drop to 2,949,111.42. This tendency is reflected in the upper and lower limits, which have values between 4,165,588.2 and 4,108,064.45 for the upper bound and 1,890,512.83 to 1,790,158.39 for the lower bound. This means that the sales forecast for Store 25 will continue to drop.

Figure 9: Visualization of HOLT-WINTERS test for STORE 20 sales

Figure 10: Result of HOLT-WINTERS test for STORE 20 sales

The sales data from STORE 20 were smoothed using the HOLT-WINTERS exponential projection for five periods. The predicted values show an upward trend over the specified periods, rising from 9,692,132.56 to 9,838,792.22. The forecast's upper and lower ranges are also climbing, with upper bounds falling between 12,274,556.54 and 12,428,330.21 and lower bounds between 7,109,708.57 and 7,249,254.23 in size. This implies a steady upward growth trajectory for the forecast's accuracy for sales at STORE 20.

Figure 11: Visualization of ARIMA test for STORE 25 sales

Figure 12: Visualization of ARIMA test for STORE 20 sales

Figure 13: Quality performance of ARIMA model for STORE 25 sales

 

The quality performance of the ARIMA model for STORE 25 sales is encouraging. The MAE (9,455.64) and MAPE (0.0034%) are low, indicating that the forecasts are correct. Moderate variability is shown by RMSE (29,901.35). The model outperforms a naive strategy, according to MASE (0.460). The model's appropriateness is supported by its AIC and BIC values of 73,748.40.

Figure 14: Quality performance of ARIMA model for STORE 20 sales

For STORE 20 sales, the quality performance of the ARIMA model is inconsistent. RMSE (86,950.12) denotes increased variability whereas MAE (27,496.04) and MAPE (0.0033%) suggest relatively accurate predictions. MASE (0.508) indicates that the model performs somewhat better than a naive strategy. A reasonable model fit is indicated by the AIC (78,652.94) and BIC (78,658.86).

Figure 15: Quality performance of HOLT-WINTERS test for STORE 25 sales

The performance of the HOLT-WINTERS model for STORE 25 sales contains flaws. A bias is evident from the Mean Error (ME) value of -37,486.18. Despite having moderate RMSE (580,387.03) and MAE (435,527.36) values, MAPE (15.47%) indicates significant percentage errors. The positive MASE (0.708) denotes relative improvement, while the negative ACF1 (-0.097) suggests that the predictive model may have been overfitted.

Figure 16: Quality performance of HOLT-WINTERS test for STORE 20 sales

The performance of the HOLT-WINTERS model for sales at STORE 20 shows limitations. The Mean Error (ME) of -152,449.83 shows that forecasts are biased. MAPE (13.54%) and MASE (0.731) point to accuracy issues while RMSE (1,317,587.47) and MAE (1,043,392.14) show substantial mistakes. The low ACF1 (-0.25) suggests that the prediction model may have been overfit.

Key Findings and Recommendations

Key Findings:

1. STORE 20 frequently outsells STORE 25, especially throughout the winter.

2. Holt-Winters forecasting works well for STORE 20 because of its ascending trend, but ARIMA works well for STORE 25 because of its declining pattern.

Recommendations:

1. In order to capitalise on the increasing trend, Holt-Winters will be useful for STORE 20 sales estimates.

2. In order to consider into account its decreasing tendency, ARIMA will be used for STORE 25 sales predictions.

3. Strategic resource allocation will be advantageous to maximise sales for each shop based on its unique trends.

Conclusion

Precision and strategic planning are greatly improved by data-driven forecasting and decision-making. We visualised and examined sales trends for STORE 20 and STORE 25 using a variety of forecasting models, including ARIMA and HOLT-WINTERS. The findings offer guidance for developing tactics that can take advantage of the unique sales trends found in each location.
 

Bibliography

 

Fill the form to continue reading

Download Samples PDF

Related Sample

Assignment Services