Date created

2020-07-27

Project Title

Alteryx Projects

Description

This project contains several projects using Alteryx,Tabluea, and Excel to do business analysis.

Projects

Project 1.1 Use Linear Regression in Excel

• Use Linear Regression to predict Diamond price with visualization in Excel, based upon previous data as predictor variables.

Project 1.2 Use Logistic Regression in Alteryx

• Choose from several numerical and categorical predictor variables by testing linear relationship.
• Used P-value to test statistical significance. Used scatter plots to check trends.
• Create a Logistic Regression model with Alteryx and justify it with Adjusted R-Square Values.

Project 4 Classification

• For this binary classification project, need to classify customers into two types based upon previous data.
• Separate training set, cleaned the data(missing data imputation, outliers, low variability), check if there is highly-correlated variables; check scatterplot after imputation.
• Used Logistic Regression, Decision Tree, Forest Model and Boosted Tree Model to create models and compare them to pick the best.
• Use Alteryx to create models and compare models with validation data, like overall accuracy, confusion matrix, and ROC curve (receiver operating characteristic).

Project 5 A/B Test a Market Dicision

• Match treatment and control units from identifying control variables and checking the correlation between control variables and performance metric.
• Use Alteryx to create the model and evaluate the results by check lift and significant level of T-test.

Project 6 Forecasting Sales with Time-based Regressions

• Decide holdout sample size based upon continuous time interval.
• Determine Trend, Seasonal, and Error components (additive or multiplicative) using time series plots.
• Tried several ETS models with Alteryx with different p,d,q settings.
• Tried several ARIMA models with Alteryx with different p,d,q and P, D, Q settings by plotting ACF and PACF to show the Time Series and Seasonal Difference after 1 or 2 differencing.
• Use AIC(Akaike Information Criteria) and calculated errors, like RMSE and MASE (below the generic 1.0) to compare the models and pick the better ones.
• Forecast with the models. Check the forecast error measurement against the holdout sample. Use calculated error measurement, choose the best model and forecast the periods.
• Graph showing 95% and 80% confidence interval.

Project 7 Combine Segmentation(Clustering), Non-binary Classification, and Time-based forecasting

• Use old store data to set up segmentation and use classification methods to classify new stores.
• Decide optimal number of store formats based upon K-Means Cluster measurement like AR(Adjusted Rand Index) and CH(Calinski-Harabasz index) methods.
• Use Alteryx to create K-Centroid Model to cluster the stores.
• Use Tableau to show the locations of clustered stores on map.
• Test Classification methods, like Decistion Tree, Forest Model and Boosted Model to classify new stores.
• Test and choose the best-fit time-based forecasting models using ETS and ARIMA algorithms.
• Apply the best model to predict future production.
• Use Tableau to demonstrate the historical data, existing stores forecasts and new stores forecasts.

Files listed

• PDF reports or the Projects
• .ymxd files need Alteryx to open

Software used

• Alteryx
• Excel
• Atom
• Tableau

Credits

• Those projects are assignments of Udacity Business Analyst Nanodegree Program.