| 摘要 |
This invention comprises a method and system for estimating unique visitors for Internet sites that is more accurate than the existing unique cookie/unique address counting methods. The invented method relies on the count of unique user identifiers (such as network addresses or preferably cookies)—I—that can be obtained from an existing cookie tracking/user access logging system. The number of unique visitors U is calculated substantially as a ratio of the count of unique cookies (or unique network addresses) to the number of visits N times the inflation factor X plus constant on that is approximately one (exactly one in the case of cookies). The number of visits is calculated by multiplying the sampling period t to the visitation frequency T1 minus one. The resulting estimate of the unique visitors is stable and does not diverge with sampling time unlike estimates directly obtained from the unique network address or unique cookie counts. The method is also applicable when there are multiple dominant visitation frequencies by accounting to the sum by all significant visitation frequencies. All key parameters of the method can be established before hand by mining a multitude of the site's historical visit logs and/or third party site access logs; the parameters can be corrected/calculated dynamically by mining the site's current access log (or current third party logs) while focusing on unambiguously identified visitors (such as return visitors identified by their login ID or unchanged cookie value).
|
