Final+Report+-+Don't+Get+Kicked!

=**DON'T BE KICKED** =

 By:
 * ===Hussain Al Haddad ===
 * ===Nathaniel Giscombe ===
 * ===Karina Javier ===
 * ===Megha Lingajappa Satish Kumar ===

=**1. Executive Summary of Findings** = Kicked cars are autos bought in an auction that have mechanical problems or any other issues that prevent it from being sold to customers, and present a loss for dealerships. The better dealerships can predict whether they are purchasing a possible kicked car or not, the better they can minimize their losses and maximize profit. The goal of this project is to develop a prediction model that dealerships can utilize when shopping for cars at auctions. This model should enable dealerships to minimize the risk of buying kicked cars by predicting whether a car is kicked or not based on some attributes. = = In the modeling process, we utilized data available at www.kaggle.com that contains around 70,000 rows of data. The target variable is “IsBadBuy” which indicates whether a car was a kick (bad buy) or not. There are 32 predictors which included information about the specifications of the car, information specific to the auction, and information about price/cost measures. = = The file contained redundant data, poor quality variables, and NULL fields for several predictors. The data was prepared and cleaned before starting the analysis. We used the following techniques to find out what works:
 * Auto Classifier
 *  Auto Cluster
 *  Cut-Off Value Adjustment
 *  Split Models
 *  Discriminant Analysis
 *  Decision Trees
 *  Dimension Reduction Techniques

Some models were not very helpful like clustering. Other models were almost identical in terms of output. We used the auto classifier node as our base technique to which we compared results of other techniques. The model with the best results was discriminant analysis which showed the following predictors as the most important 1) “VehicleAge”, 2) “MMRCurrentAuctionCleanPrice”, 3) “MMRCurrentRetailAveragePrice”, and 4) “WarrantyCost”.

=
The model achieved a percentage of correctly predicted good buys at 91.75% which can save huge amounts of wasted money for car dealerships. 91.75% is 4.05 percents higher than 87.7% achieved by random/subjective car selection. This improvement can result into huge savings that range from thousands to millions of dollars depending on the volume of sales the model is applied to (e.g. $607,500 for 10,000 used cars sold).======

=
To improve these results, it would be interesting to add a variable that shows past accidents related to the cars and see how it affects the probability of an auction car to be kicked. A vehicle reliability rating where customers gave input on how satisfied they were with the car would also be a useful variable for this model. ======

=<span style="font-family: Arial,Helvetica,sans-serif;">**2. Background** = The used automobile market represents nearly half of the retail automobile sales market in the US. This industry generates about $370 billion in annual sales, making it one of the largest retail segments in the US economy. Most car dealerships acquire used automobiles by either trade-in or auction purchases. = = For automobile dealers, used automobiles represent a large portion of revenue and since the margins in this industry aren’t very high, about 20% on average, it’s important for dealers to sell large quantities. For example, in 2011 44 million used cars sold in the US alone, compared to 17 million new cars. = = Purchasing used cars from auto auctions is a risky business. It is one of the biggest challenges an auto dealership can face due to the risk of the vehicle having serious issues that prevent it from being sold to customers. The auto community calls these unfortunate purchases "kicks". = = Kicked cars often result when there are tampered odometers, serious mechanical problems, issues with getting the vehicle title from the seller, or some other unforeseen problems. Associated costs include transportation cost, throw-away repair work, holding costs, and market losses in reselling the vehicle. Kicks are costly to dealerships whether they wanted to fix their issues or sell it as-is. = = Dealerships goal is to provide the best inventory selection possible to their customers otherwise they lose reputation, customers, and profitability. A model that help figure out which cars have a higher risk of being kick can provide enormous value to dealerships. By reducing the probability of purchasing a kicked car, dealers would be able to provide better inventory for customers, minimize their costs, and in turn increase their profits. = = To illustrate the significance of the problem consider the following calculations:
 * <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: center;">** Item ** || <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: center;">** Amount ** || <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: center;">** Notes ** ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;">Number of used cars bought & sold by a dealership || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: right;">10,000 || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;"> Hypothetical ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;">Percentage of kicked cars || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: right;">12.30% || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;"> According to the data-set ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;">Number of kicked cars || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: right;">1,230 || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 90%;"> Number of used cars * Number of Kicked Cars ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">Average Price per used car sold || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: right;">$ 10,000 || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> Assuming $10,000 average price for used cars. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">Profit on a normal car || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: right;">$ 2,000 || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> 20% = $2,000 is average profit margin. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">Profit on a kicked car || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: right;">$ 500 || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> Assuming Profit margin drops to $500 in case of a kicked car. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">Loss of potential profit || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: right;">$ 1,500 ||  ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">Total loss of potential profits || <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: right;">$ 1,845,000 ||  ||

$1,845,000 is a lot of money that our dealership is losing. Is there a way we can save some of this money? Yes, this is why we are working on this project. How much can we save from the wasted potential profits if we create the model? Well, it depends on the reliability of the available data. The following table shows possible savings for each percentage of improvement (lift) over random (subjective) selection of cars:
 * <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">** Additional profits from each % **** Of Improvement (lift) ** ||
 * <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">1% || ======<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> $ 150,000 ====== ||
 * <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">2% || ======<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> $ 300,000 ====== ||
 * <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">3% || ======<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> $ 450,000 ====== ||
 * <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">4% || ======<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> $ 600,000 ====== ||
 * <span style="color: #000000; display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">5% || ======<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;"> $ 750,000 ====== ||

Note that the impact of the improvement resulting from the model is relative to the volume of sales. As the exposed volume of sales becomes larger, the profits resulting from the model are expected to grow accordingly. If we apply the formula in the table above to the whole industry (44 million used cars) it accumulates to $ 66 billion loss of potential profits. This translates into $660 million of profits driven by each percentage of improvement due to the model.

=**3. The Problem**=

To predict if the car purchased at the Auction is a Kick (bad buy).

=<span style="font-family: Arial,Helvetica,sans-serif;">**4. Data Understanding** =

The data set obtained provided by Caravana via Kaggle contains more than 70k rows of records. According to the problem, the target variable is “IsBadBuy” which indicates whether a care is a good buy or not and there are 32 predictors of varying quality. As a step to data understanding, we categorized the variables into three groups: Car Specifications, Auction Specific, and Price/Cost. For a list of the variables and their descriptions refer to the Appendix. Those variables needed thorough inspection for cleaning and preparation. The following paragraphs discuss highlights of our findings.

a. Redundant Data
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> The data set contains redundant variables. “VehYear” and “VehicleAge” are two variables with almost the same meaning. We will filter “VehYear” out while modeling. It is the same case for variables “VNZIP” and “VNST” as well. We will set “VNZIP” as none instead filtering it out so it remains accessible for reference and use at later stages.

Among the variables in the data set two variables were of significantly poor quality. This first is “PrimeUnit” which identifies if the vehicle would have a higher demand than a standard purchase. Most of the records under this variable were NULL and only a small portion showed “Yes” or “No”. Thus, this variable should be excluded. <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; line-height: 0px; overflow: hidden; text-align: justify;">

<span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> The second poor quality variable was “PurchDate” which shows the date the vehicle was purchased at Auction was. Over 60% of the records in the data set did not have a purchase date recorded. In addition, we do not believe purchase date will add any value to the predictions. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

**<span style="color: #17365d; font-family: Arial,Helvetica,sans-serif;">c. Other Observations **
<span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> We inspected some of the variables that we think are major for concentration of a certain value. Starting with the target “IsBadBuy”, more than 87% of the records are 0s: <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;">

<span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> The distribution of “Auction” among the three auction providers is not bad. Also, the proportion of good and bad buys for cars from each auction seems to be normal. <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> Under the “Transmission” variable the value “Auto” prevails in more than 96% of the records. This has a serious negative impact on the usability of the variable. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> Some other data set issues include: <span style="font-family: Arial,Helvetica,sans-serif; font-size: 90%;">
 * ”AucGuart” should have three categories: Green, Yellow, and Red. In the data set there were Green, Red, and NULL. So, we assumed NULL is yellow. Later we noticed that NULL (Yellow) composed over 95% of the data which does not very much help the model recognize the effect of each of the three values in the variables.


 * Under the variable “Transmission” we needed to reclassify “manual” (lower case) as MANUAL (upper case) because SPSS Modeler was case sensitive when in reading records which raised a concern during analysis.

=**5. The Stream**= <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> When we started developing our stream, after doing the initial data auditing and maintenance steps, we wanted to explore all possible avenues. So, in order to be more efficient and use our limited resources wisely, we decided to use automated modeling nodes: Auto-classifier and Auto-cluster. <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;"> According to the insights we derived from the automated modeling nodes and our understanding for the data, we expanded in the data mining process looking for more valuable insights. Our ultimate goal was always brining value to dealerships and the problem of our project was always at the core of our concern. The following visuals of the 2 streams we built tell a summary of our data mining story:

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

=
<span style="font-family: Arial,Helvetica,sans-serif; font-size: 90%;">The following section will display insights derived from on our data mining efforts. It will also highlight successes and failures among our efforts to obtain valuable information. ======

=**6. Insights and Highlights**=

In order to have a summary of the models that would work best in making a prediction we ran auto-classifier node. The results we received after running the auto classifier showed logistic regression with 89.5% accuracy, neural net with 89.6% accuracy, and CHAID tree with 89.7% accuracy. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> Logistic regression helped us better understand each symbolic field in the model and the sample count within each variable by breaking down the records in the training data. Based on all three models we saw that Wheel Type showed up as one of the most important predictors. Neural net showed how each variable is connected to the output highlighting the strongest connections. Inputs with strongest connections are: vehicle age, vehicle acquisition cost (VehBCost), and wheel type. CHAID tree shows wheel type ID leading the tree and warranty cost, vehicle age, MMR acquisition auction clean price, and auction provider as final predictors. In the auto classifier we also observed that the Logistic model had the highest lift value compared to the other two models, meaning the records in that node were more likely to fall under the target category.

** b. Adjusted Cut-Off Value to Maximize Profits: **
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> We’re going to assume that dealerships make a profit of about $2,500 on each good car sold, while they make about $500 on each kicked car sold. In order to mitigate the risks of purchasing a kicked car and losing $2,000 of potential gain, we have decided to implement a cut off value of 0.8. This cut-off value was derived using the table below:

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> Based on the tables above we can see that using a standard cut off value of 0.5 we get 90.3 % correctly predicted good cars compared to 91.5 % by adjusting the cut-off to 0.8. This is a small increase but we are lowering our probability of purchasing a kicked car by increasing our accuracy for determining good cars. As a result, we can conclude that profitability is increased. The impact of this increase in profitability is proportional to business volume.

**<span style="color: #17365d; font-family: Arial,Helvetica,sans-serif;">c. Split Models **
<span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;">To verify whether we can improve the accuracy of the model, we split the data set based on the “AUCGUART” variable into 3 splits: GREEN, RED and YELLOW (NULL). “AUCGUART” is the level of guarantee on cars provided by the auction. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;">Based on a unique classification model for each split, the prediction accuracy for the GREEN, YELLOW, and RED splits varied from them base model. The highest improvement is by 6% in the GREEN split. Clearly, the predictions for the YELLOW split are less accurate. This is probably due to the fact that a wide range of cars conditions are classified as YELLOW since it is the middle class. Indeed, the vast majority of the records are classified as Yellow (NULL). <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> <span style="display: block; font-family: Arial,Helvetica,sans-serif; font-size: 90%; text-align: justify;">We also split the data based on the “Auction” variable into 3 splits: ADESA, MANHEIM, and OTHER. “Auction” identifies the auction provider at which the vehicle was purchased. The results of the split models compared to the base/general model were as follows: <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">Again, the largest category achieves the least accurate model. This is probably due to the variability resulting from split size or because over-fitting is easier for smaller (less varying) splits. However, the difference is not big enough to raise a critical concern.
 * **Splits Based on “AUCGUART”:**
 * **Splits Based on “Auction”:**

<span style="font-family: Arial,Helvetica,sans-serif;">The discriminant analysis helped us identify the important predictors. The output of this analysis is as shown:
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> We see that “VehicleAge”, “MMRCurrentAuctionCleanPrice”, “MMRCurrentRetailAveragePrice” and “WarrantyCost” are the most important predictors for classifying vehicles as bad buy or good buy. When the “VehicleAge” is less we can expect the car to be a good buy than of a car with a bigger Vehicle Age. We can also say that on the day of auction, the acquisition price of the vehicle is also an important predictor as the managers of the auction might want to sell the vehicle at a higher price. The following classification matrix derived from the Discriminant Analysis shows how accurate the model is in predictions: <span style="display: block; font-family: Arial,Helvetica,sans-serif; line-height: 0px; overflow: hidden; text-align: justify;">

The most important part of it is the percentage of correctly predicted good buys at 91.75%. Because purchase decisions are based on the predictions of good buys our focus is to maximize the percentage of correctly predicted good buys. On the other hand, only 8.24% of actually bad cars are were predicted to be good cars based on the model.

** e. Decision Tree **
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">The most important variable appeared to be vehicle age. For cars older than 6 years there is a 76% probability that they will be a good buy and 23% probability that it will be a bad buy. Cars between 4 and 6 years and Alloy wheel type, which are used to improve cosmetic appearance, have 86% probability of being good buy and 13% bad buy. Cars that contain covers have 88%probability of being good buy and 11% bad buy. However, cars that do not contain alloy or covers have 60% probability of being a bad buy and 40% good buy.Cars that are less than 4 years old have an average of 57% probability of being a bad buy if they do not have wheel covers or alloy wheels.

=
<span style="font-family: Arial,Helvetica,sans-serif;">We tried several clustering techniques including the auto cluster node. Unfortunately, we were not able to create a model with good clustering quality. All of the models were poor with absolute silhouette values less than 0.2 each. ====== <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

=
<span style="font-family: Arial,Helvetica,sans-serif;">Of course, considering the quality mentioned earlier, the bagged model could not separate the records into clusters effectively on a plot using the 2 most important predictors. ====== <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

<span style="font-family: Arial,Helvetica,sans-serif;">Also, the clusters did not seem to separate records of bad buys from those which are not.
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">

** g. Dimension Reduction Techniques **
Feature Select identified all predictors as highly important; it did not help reduce the number of variables in the data set.
 * Feature Select**

The PCA produced 3 components which we used 2 of them to replace variables that were highly correlated with the components. The following charts show the components, correlations, and the replaced variables. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> <span style="font-family: Arial,Helvetica,sans-serif;">After using the PCA technique on the data set, no significant changes were noticed. Wheel Type is still an important predictor followed by the 2 added PCA factors. However, models’ accuracy neither improved in predictions nor in clustering with PCA.
 * PCA**

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;"> =<span style="color: #000000; font-family: Arial,Helvetica,sans-serif;">**7. Conclusions** =

** b. Best Model: **
<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">The best model was the discriminant analysis model. It listed the following predictors as the most important ones for classifying vehicles as bad buy or good buy:1) “VehicleAge”,2) “MMRCurrentAuctionCleanPrice”,3) “MMRCurrentRetailAveragePrice”, and4) “WarrantyCost”. <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">The model also showed slightly higher accuracy over the auto classifier technique we used. The most important part of it is the percentage of correctly predicted good buys at 91.75% compared to 90.35% for the auto classifier. This slight improvement can be translated into hundreds of thousands of dollars.91.75% is higher than 87.7% achieved by random/subjective car selection. This makes a 4.05 difference. To measure how much difference this improvement can make for dealerships in terms of dollars, we will use the following table (for the complete table refer to the appendix): <span style="display: block; font-family: Arial,Helvetica,sans-serif; line-height: 0px; overflow: hidden; text-align: justify;">

** c. Suggestions for Project Extension **
In addition to the variables provided by kaggle, we thought that there were some other variables that could have been included in the data set that could have helped the accuracy of our model. For example, we could have used a variable similar to a car fax report that would reveal whether a certain vehicle was involved in an accident or had a history of water damage. Vehicles with a history of damage could potentially become a kicker due to the fact they already have a history of problems. Another variable that could have increased the accuracy of our model would be a vehicle reliability rating from J.D. Power and Associates. J.D. Power and Associates is a marketing information service company that provides customer satisfaction research, market research, automotive forecasting as well as other performance rating programs. This reliability rating variable could potentially identified vehicles with a history of being unreliable therefore increasing the probability of identifying a kicked vehicle. A significant matter is the cost of obtaining the data for the two variables. This is likely to hinder future efforts for model improvement. An additional effort for a better that can be done is collecting data of better quality. In the current dataset, there were several predictors with poor quality as we emphasized earlier in this report.

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">_

<span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">**LIST OF VARIABLES**
 * <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">** Auction ** ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">AcquisitionType || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Identifies how the vehicle was acquired (Auction buy, Trade in, etc). ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">AUCGUART || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The level guarantee provided by auction for the vehicle (Green light - Guaranteed/arbitrable, Yellow Light - caution/issue, red light - sold as is). ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Auction || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Auction provider at which the vehicle was purchased. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">BYRNO || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Unique number assigned to the buyer that purchased the vehicle. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">KickDate || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Date the vehicle was kicked back to the auction. ||
 * <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">** Car Specification ** ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Color || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicle Color. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">IsBadBuy || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Identifies if the kicked vehicle was an avoidable purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">IsOnlineSale || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Identifies if the vehicle was originally purchased online. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Make || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicle Manufacturer. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Model || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicle Model. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Nationality || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The Manufacturer's country. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">PRIMEUNIT || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Identifies if the vehicle would have a higher demand than a standard purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">PurchDate || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The date the vehicle was purchased at Auction. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">RefID || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Unique number assigned to vehicles. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Size || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The size category of the vehicle (Compact, SUV, etc.) ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">SubModel || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicle sub model. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">TopThreeAmericanName || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Identifies if the manufacturer is one of the top three American manufacturers. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Transmission || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicles transmission type (Automatic, Manual) ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">Trim || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Vehicle Trim Level. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VehicleAge || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The Years elapsed since the manufacturer's year. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VehOdo || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The vehicles odometer reading. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VehYear || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The manufacturer's year of the vehicle. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VNST || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> State where the car was purchased. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VNZIP || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Zip code where the car was purchased. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">WheelType || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The vehicle wheel type description (Alloy, Covers) ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">WheelTypeID || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> The type id of the vehicle wheel. ||
 * <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: center;">** Price/Cost ** ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRAcquisitionAuctionAveragePrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in average condition at time of purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRAcquisitionAuctionCleanPrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the above Average condition at time of purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRAcquisitionRetailAveragePrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the retail market in average condition at time of purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRAcquisitonRetailCleanPrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the retail market in above average condition at time of purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRCurrentAuctionAveragePrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in average condition as of current day. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRCurrentAuctionCleanPrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the above condition as of current day. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRCurrentRetailAveragePrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the retail market in average condition as of current day. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">MMRCurrentRetailCleanPrice || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition price for this vehicle in the retail market in above average condition as of current day. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">VehBCost || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Acquisition cost paid for the vehicle at time of purchase. ||
 * <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;">WarrantyCost || <span style="color: #000000; font-family: Arial,Helvetica,sans-serif; font-size: 10.66px;"> Warranty price ||

<span style="font-family: Arial,Helvetica,sans-serif;">**SIGNIFICANCE CALCULATIONS** <span style="display: block; font-family: Arial,Helvetica,sans-serif; text-align: justify;">