Information Intelligence Corporation
Toronto, ON
Company Overview
NO single analytical software package in todays marketplace will provide a complete solution to your data mining and system modeling needs. Most companies requiring sophisticated data analysis use several technologies to achieve their desired results.
Unfortunately most business data is approximate or uncertain yet the mathematical techniques commonly used assume that numbers are exact. To deal with this incorrect assumption, we use statistical techniques based upon distributions of many of these exact numbers. Most statistical techniques work properly with only a limited range of different distributions. By giving up the dependency upon exactness, IIC fuzzy techniques more closely model the real situation and can produce significantly better results in the magnitude of 10% or greater.
SEE BELOW SOME OF IICs SUCCESS STORIES
Leading Canadian Financial Institution works with IICs advanced fuzzy engine
Financial services companies such as banks and credit unions routinely investigate the creditworthiness of their clients in order to determine what services may be offered to them. A case in point is determining which type of credit card is appropriate to the clients financial situation. Banks are interested in determining the risk adjusted return to the bank given the credit card product offered to the client.
In a series of prototype system modeling experiments IIC determined the 'risk adjusted return' for a set of 900+ clients.. Inputs included such variables as: checking balance, saving balance, income bracket, mortgage, etc. With such input and output variables, IIC built a fuzzy rule base with both unsupervised and supervised learning to predict potential risk adjusted return of a client based on his/her particular variable values.
We identified low, medium and high levels of 'risk adjusted returns' for the bank with a high level of accuracy as compared to the 'logistic-curve' analysis that they are currently applying.
Other areas within the organization have been identified that can benefit from the IIC fuzzy engine include
Relationship Marketing (householding) Customer Profiles Direct Marketing Marketing, Campaign Management Client Retention Risk Management including Credit Monitoring:
Fraud and delinquency triggers Predictive Analysis to increase profits from credit card holders IIC works with World Renown Internet Technology Leader Current practice among Internet Service Providers is to support their customers with Best Effort level of service. Many ISPs have Service Level Agreements that determine the Quality of Service based on the Class of Service that was purchased. Failure to meet the terms of the SLA may impose financial penalties on the ISP based on any change in the level of service that is provided to the customer. Agreements tend to be divided into three classes of service: Gold, Silver, and Bronze. Examples would be Gold for video on demand, Silver for voice and Bronze for text data transmission.
ISPs, Network Equipment, and Network Software Suppliers are seeking ways to increase their revenue and provide improved customer service by better managing network resources through the use of a network traffic flow decision support system. Such systems must be able to predict network traffic demand and allocate sufficient bandwidth for both backbone and network nodes. Dynamic adjustment of bandwidth allocation must occur within microseconds in order to alleviate network congestion. Current practices see bandwidth allocation being manually adjusted by network personnel during low utilization periods
IIC has demonstrated the ability to create fuzzy rule base models that can accurately predict traffic demand and allocate bandwidth dependant on the Class of Service to be provided to the customer.
The bandwidth allocation can be done within a Virtual Path if the IP provider does not want to invest in extra capital equipment. ISPs willing to investigate technologies such as wavelength switching at the nodes could be provided with bandwidth allocation schemas between virtual paths. Local Healthcare Researchers to more accurately determine drug dosages
In pharmacology, analysts are interested in determining dosage level of medication to individuals based on a person's particular attributes such as age, weight, etc. For example recently, its become clear that all adults should not be given the same dosage. Their age, weight and other attributes should be taken into account in order to minimizing side effects and induce better treatment. Currently, IIC is conducting comparison experiments for lithium dose and serum concentration prediction between fuzzy system modeling approaches and multiple linear regression approaches. The input variables are level, sex, age, weight, status and tetra cyclic antidepressants which determine the dose of daily lithium carbonate in mg's where level is serum lithium level in mmol/L.
IIC improves predictive capability for Continuous casting Customer orders come in for various grades of steel, which specify width, length, weight, due date, etc. If the molten metal pored into the ladle of a continuous caster is changed frequently in order to deliver the customer orders without delay, a large amount of unspecified grade of steel is produced between different customer orders which require distinct grades of steel.
IIC has developed a model that minimizes the tardiness of customer order delivery due dates and the amount of mix grade (unwanted) steel produced between orders. When compared with a multi-variable regression model, our fuzzy system model gave a better prediction with a reduction of error at 11% level, i. e., multi-variable regression model gave %15
The IIC fuzzy engine is based upon more than 20 years of research in the area of fuzzy logic application by Dr. Burhan Tursken, a world leader in this field and current chief scientific officer at IIC.
Company Information
Physical Address
.
Toronto, ON M5S 1R8
CA
Mailing Address
123 Bloor St W, Suite 200-211
Toronto, ON M5S 1R8
CA
Additional Details
Legal Name: Information Intelligence Corporation
Number of Employees: 11
CA
Data Mining and System Modeling Processes
The Ten Steps of Data Mining and System Modeling Presented is the IIC process for extracting hidden knowledge from a data warehouse, a customer information file, or a company database.
- Identify The Objective
Before we begin, you have to be clear on what you hope to accomplish with our analysis. We need to know in advance the business goal of the data mining and system modeling for your company. We need to establish whether or not the goals are measurable. Some possible goals are to:
find sales relationships between specific products or services identify specific purchasing patterns over time identify potential types of customers find product sales trends.
@ These items are some of customer related processes for data mining and systems modeling project.
- Select The Data
In working with you to help define your goals, the next step is to select the appropriate data to meet this goal. This may be a subset of your data warehouse or a data mart that contains specific product information. It may be your customer information file. We need to segment as much as possible the scope of the data to be mined.
Here are some key issues.
Is the data adequate to describe the phenomena that the data mining analysis is attempting to model? Can you enhance internal customer records with external lifestyle and demographic data? Is the data stablewill the mined attributes be the same after the analysis? If you are merging databases can you find a common field for linking them? How current and relevant is the data to the business goal?
@ Most of these (and similar) issues will probably be centered around customer related processes. IIC works with the data mining project manager and the IT personnel to decide on the scope of the data to be mined.
- Prepare The Data
Once the data has been assembled, a decision must be made as to which attributes to convert into usable formats. Consider the input of domain expertscreators and users of the data.
Establish strategies for handling missing data, extraneous noise, and outliers Identify redundant variables in the dataset and decide which fields to exclude Decide on a log or square transformation, if necessary Visually inspect the dataset to get a feel for the database Determine the distribution frequencies of the data
- Audit The Data
Evaluate the structure of data in order to determine the appropriate tools. What is the ratio of categorical/binary attributes in the database? What is the nature and structure of the database? What is the overall condition of the dataset? What is the distribution of the dataset?
We balance the objective assessment of the structure of data against users' need to understand the findings. Neural nets, for example, don't explain their results but fuzzy rule-bases provide patterns and linguistic insight.
- Select The Tools
Two concerns drive the selection of the appropriate data mining and system modeling tools; a companys business objectives and its data structures. Both should guide us in the selection of the tools.
Consider these questions when evaluating a set of potential tools. Is the data set heavily categorical? What platforms do candidate tools support? What data format can the tools import?
No single tool is likely to provide the answer to a data mining and system modeling project. Some tools integrate several technologies into a suite of statistical analysis programs, a neural network, and a symbolic classifier or fuzzy pattern recognizer.
@ The IIC fuzzy system modeling approaches provides a universal approximator, which is based on fuzzy rules. The extracted fuzzy rules give an insight of the data, and the inference provides an estimate for the business decisions. The extraction of the rules is achieved by numeric fuzzy clustering methods such as FCM and PCM. The clusters are converted to fuzzy rule-base by a rule- base- formation module. The display module converts them to easy- to-understand plots. The inference modules accomplish the fuzzy reasoning of the new data, and gives out the estimated value of the concerned feature. The parameter optimization and tuning modules provide the selection and correction of many inference parameters.
- Format The Solution
The output of your data audit will depend on the business objectives and tool selection made. The Key questions are:
What is the optimum format of the solutiondecision tree, rules, C code, SQL syntax? What are the available format options? What is the goal of the solution? What do the end-users needgraphs, reports,
IICs currently developed and implemented approach provides estimates for business decision support through the inference module, and rule-base graphs for human understanding of the rules through a display module.
- Construct The Model
The data mining process begins at this point. The first step typically is to use a random number seed to split the data into a training set and a test set and construct and evaluate a prototype model. The generation of classification rules, decision trees, clustering sub-groups, scores, code, weights and evaluation data/error rates takes place at this stage. The following issues need to be resolved at this point:
Are error rates at acceptable levels? Can we improve them? What extraneous attributes did we find? Can we purge them? Is additional data or a different methodology necessary? Will we have to train and test a new data set?
@ IICs currently implemented approach learns the rules from a training data set, and verifies using a test data set. The implemented modeling tools are based on the state of the art approaches, and they are open to improvement through research- development cycles.
- Validate The Findings
Share and discuss the results of the analysis with the business client or domain expert. Ensure that the findings are correct and appropriate to the business objectives. Do the findings make sense? Do we have to return to any prior steps to improve results? Can you use other data mining tools to replicate the findings?
@ Currently developed modeling tools accommodate various error criteria to provide measures of fitness and to observe the improvements of the target rule-base. Depending on the largeness of the data, and the simplicity of the rules, there are several alternate tools such as multivariate least square regression analysis and similar statistical tools that are incorporated into our fuzzy system modeling.
- Deliver The Findings
Provide a final report to the business unit or client. The report should document the entire data mining process including data preparation, tools used, test results, and rules. Some of the issues are:
Will additional data improve the analysis? What strategic insight did we discover and how is it applicable? What proposals can result from the data mining and system modeling analysis? Do the findings meet the business objective?
@ An expert must combine the outputs of our tools to a business report that points to the listed issues. An expert is necessary for the final decision of the validity and applicability of the result to the business.
- Integrate The Solution
The findings should be presented to all interested end-users in the appropriate business units, as the incorporation of results could necessitate the alteration of a company's business procedures. Some of the data mining solutions may involve
SQL syntax for distribution to end-users C code incorporated into a production system Rules integrated into a decision support system.
Although data mining and system modeling tools automate database analysis, they can lead to faulty findings and erroneous conclusions if one is not careful.
Bear in mind that data mining is a business process with a specific goalto extract a competitive insight from historical records in a database.
Contacts
Products
- Fuzzy Logic Data Engine
Additional Information
541510