Patrick Duff Analytical Algorithm Whitepaper Introduction This paper is directed to small business owners desiring to use analytical algorithms in order to improve sales, reduce attrition rates raise profits and reduce wasted capital. How can this be done? This can be done via predictive analytics, which this paper will focus on. The information will be taken from the Microsoft Developer Network. While the content focuses on the Microsoft algorithms, the underpinnings will apply regardless of the manufacturer. Predictive analytics are being used to assist in making better decisions through the scientific analysis of consumer behaviors. Current fields that utilize predictive analytics that will be talked about below include but are not limited to: marketing, sales and finance. Because these algorithms are highly adaptable they are being used in more and more fields. Background The essential function of analytics is to attempt to predict a consumer s future purchases based upon previous purchases. The heart of analytics is an algorithm, and these algorithms are designed to carry out different types analyses to provide different types of information. While there are many algorithms, this white paper will focus on two: regression and clustering algorithms. These two have been chosen because
they are the most basic and can be some of the most useful in gathering data that will allow much more accurate adverts, predict the success of these adverts and a clients needs and wants, all based on prior interactions with a website. About the algorithms: 1 Clustering algorithms: predict one or more discrete variables, based on the other attributes in the dataset. In the simplest terms it groups like variables together. can be used for: grouping current and potential clients together by age, location, income, etc. for targeted ad campaigns Aiding in determining the best market for an initial product release. Clustering can be used to group all the shopping items available on the web into a set of unique products. For example, all the items on ebay can be grouped into unique products. Regression algorithms: predict one or more continuous variables, such as profit or loss, based on other attributes in the dataset. can be used for: flagging the customers in a prospective buyers list as good or poor prospects. predicting overall lifetime profit amount. When a customer needs refill of frequently purchased items. Which items a customer will most likely purchase in an upsell or cross sell. 1 Taken from: https://msdn.microsoft.com/en-us/library/ms175595.aspx
Clustering Algorithms are particularly important in identifying in surveying the landscape of the market(s) in which your products reside. What are the demographics with respect to age and income? Which groups have tastes in similar products when ignoring age? What are the outliers? Once the data that you feel is the most important are gathered the algorithm will sort through it and group like variables together. A scatter plot graph is best to see this. Regressive algorithms work by gathering a customers purchase history then performing a regression analysis on the variables to extrapolate data such as purchase intervals or inventory items with traits that have commonalities with previous purchases. As with most data modeling techniques a certain amount of statistical assumptions need to be made until enough data is gained to provide an accurate picture and regression. Data required to use clustering algorithms & regression algorithms: A single key column: Each model must contain one numeric or text column that uniquely identifies each record. Compound keys are not allowed. Input columns: Each model must contain at least one input column that contains the values that are used to build the clusters. You can have as many input columns as you want, but depending on the number of values in each column, the addition of extra columns can increase the time it takes to train the model.
The data that is required is similar to a primary key in MySQL, and serves the same function: a unique identifier for each data set. In essence these two, and every other predictive algorithm queries an RDB to do its job. The differences between the two is how the algorithm processes the db data. Benefits to using analytical algorithms: Easier to predict future levels of inventory. Less wasted time and money on ineffective product launches, advert campaigns. able to suggest products that customers will be the most likely to purchase. accuracy increases with increased usages, allowing gradually. if integrated with the cloud, data analysis can occur almost in real time, drastically increasing prediction accuracy and speed. Cons to using analytical algorithms: The algorithms that are required can require a substantial investment, which might not be feasible. The data that has been mined from client usage of your web presence needs to be used responsibly as well as stored securely which requires an investment in a security suite of software, hardware and personnel. Ideally it should be coupled with cloud computing, which unless preexisting can pose a large investment. The suite of analytical algorithms can be difficult to navigate. This requires a bit of research on which tools are right for your firm at the time of investment. Summary
The use of analytical algorithms can require a moderate investment on the business owners part, however they can and often do yield much needed information on emerging and current markets, the success rate of an entire ad campaign or certain aspects of said campaign, and other invaluable business intelligence. This information can be the impetus that your business needs to grow exponentially, provided the information gathered is used and stored responsibility. As with any major investment, a cost-benefit analysis should be undertaken along with a moderate amount of research to figure out which algorithms are correct for your situation and the future situation you wish to see yourself in. Further Reading: "Data Mining Algorithms (Analysis Services - Data Mining)." Data Mining Algorithms (Analysis Services - Data Mining). Microsoft. Web. 22 Apr. 2015. <https://msdn.microsoft.com/en-us/library/ms175595.aspx>. "Microsoft Clustering Algorithm." Microsoft Clustering Algorithm. Web. 22 Apr. 2015. <https://msdn.microsoft.com/en-us/library/ms174879.aspx>. "Microsoft Linear Regression Algorithm." Microsoft Linear Regression Algorithm. Web. 22 Apr. 2015. <https://msdn.microsoft.com/enus/library/ms174824.aspx>.
"Chapter 7: Clustering." Clustering. Oracle. Web. 22 Apr. 2015. <http://docs.oracle.com/cd/b28359_01/datamine.111/b28129/clustering. htm#i1005728>.