How does Informeta's unique and innovative software benefit you?

Informeta's artificially intelligent software engine, named "Mentys," identifies and fixes data anomalies in real time processes. The software was first conceived and scoped more than 17 years ago, before adequate hardware was available. Dr. Ron Coleman and Alan Waksman started Informeta to develop Mentys in 2001. Mentys is different from other solutions in a number of ways. While many software packages detect data errors, Mentys goes far beyond this to:

• Explain other variables that are implicated
• Propose corrections through predictions
• Learn from numeric or category data and prior knowledge
• Combine this knowledge into one prediction, and
• Give the risk level of proposed corrections/predictions.

Mentys does neither data mining nor data cleansing, but extends the data correction capabilities (compare Mentys vs. data mining or Mentys vs. data cleansing). Mentys also differs from statistical packages in a number of critical areas (compare Mentys vs. statistical packages).

Corporations are constantly required to reduce risk and improve their compliance capabilities. With ever increasing volumes of information and data anomalies, the International Financial Services Association (IFSA) predicts that financial institutions will spend 8-10% of their annual budgets in managing their risk and compliance responsibilities. The Federal Reserve and corporate shareholders are no longer accepting responses such as "I didn't know" or "It's not my responsibility," but are holding executives and employees personally responsible for their actions. Smart software that can improve timeliness and accuracy of data is now critical to avoiding investigations, penalties, and possible criminal charges.

Whom does Mentys benefit?

Informeta functions as a software OEM who sells products, support, and consulting services to end users, OEMs, and VARs.

• Banks
• Brokerage
• Insurance
• Hedge Funds
• Consulting & Software
• OEMs & VARs
• Military & Government

Current Development

Informeta is currently in the preliminary stages of developing a Letter of Credit application (L/C) for an international VAR and a major International financial end user, to identify new L/C pricing anomalies. The application requires input from three different sources of information: client historical data, external data (i.e., the Web, publications, etc.) and expert knowledge. A successful product will have major implications in satisfying the federal government requirements to trace money laundering and terrorist money transfers. The market for this specific L/C application is currently untapped. Please see our L/C white paper for more details.

Proven Successes

1. Monitoring Trading Activity

One rogue trader operating inappropriately over a short period can potentially wipe out years of company profits. It is often difficult to confirm whether illicit trades are being made, and manual audits are time-consuming and error-prone.

Using Mentys, Informeta investigated six months of foreign currency trades for a leading financial services company (see project report). By combining expert prior knowledge with data from roughly 60,000 trades, Mentys identified a specific trader whose patterns of behavior appeared highly unusual compared with his peers. Mentys also found that this trader made nearly 150 specific anomalous trades, which deviated significantly from the expected market rate.

Apart from successfully discovering evidence of unusual trading behavior, we also recognized a number of related benefits of Mentys in this area. In particular, the ability of Mentys to analyze various patterns of trade could be used to promote good trading practices and reduce poor ones.

2. Market Data Correction

Mentys was applied to find and correct simulated errors in daily market data feeds (see project report). The data used was extracted from a real "clean" file of open contracts which is used to establish appropriate reserve funds. Mentys identified more than 91% of all simulated errors and, on average, the predicted value was within 1.34% of the true value. Mentys also identified contracts that should not have been in the "clean" file because they were closed contracts. These contracts were present due to a prior system error and the client was erroneously reserving funds on $1.8 billion worth of closed contracts.

3. Credit Risk

Traders are often unable to determine, with confidence, whether or not requested trades are within the client's established credit limits. Sixty percent of the portfolios analyzed for a major financial services corporation exhibited at least one anomalous record which affected the client credit exposure during the period under investigation (see project report).

Of the sixty percent, Mentys was able to account for and correct 72% of the anomalies using available data. The remaining 28% indicate causal factors outside the data we received. With the availability of additional data, this could be accounted for, or a system error may have resulted in the fluctuation.

4. Data Recovery

Mentys successfully recreated data items for an analytics company who lost four weeks of unique data due to the 9/11 attacks on the World Trade Center (see project report). Historical data was available to aid in this recreation as well as New York Times information spanning three years, retrieved from the web. In previous work, we noted that the addition of general news information seemed to improve accuracy when filling in missing data.

In a market data example, 92.2% of Mentys' predictions were within a 5% error tolerance of the correct value using historical data alone. With news information added, 97.2% of the predictions fell within the same tolerance.

In the case of volatility data, 98.8% of predictions fell within the 5% error tolerance without news, compared to 99.8% with the news.

5. U.S. Treasuries Trading

In a simulated trading environment, Mentys was used to predict the change in percentage yield of US treasury constant maturity rates (see project report). Mentys was compared with two popular alternative software products that utilize the artificial neural network (ANN) method. The tests were run in the same trading environment with both good and bad market conditions, and showed Mentys made fewer trades with fewer highs (gains/losses), thereby reducing the risk of each trade. Using Mentys, we observed a cumulative profit of more than $300K over a six-month period. Using the competitive products we realized $40K and $20K losses.

6. Risk Rating

When Mentys identifies errors in data, it will propose the correct datum for each instance through acquired knowledge and prediction. Many business decisions require that corrections to erroneous data fall within a well-defined tolerance for error. Mentys enables you to do more than simply hope that the corrections are acceptable. With each prediction Mentys provides a reliability estimate that can be used to decide whether to accept or reject the proposal, according to your predetermined tolerance.

In one example, we used Mentys to find and correct 343 bad quotes from a corrupted feed of 837 bid and ask prices for four stocks.

By using Mentys' reliability estimates to select only the top 35% of predictions, we were able to reduce the median prediction error by more than half from 0.388 to 0.186 standard deviations.

Does Mentys benefit you?