In the statistical analysis module, developers can see how their AI capabilities are called to assist in renewal, call analysis and other related decisions.
In order to use this function module, understand the basic concepts, before understanding the specific platform operation.
Basic concept - Basic statistical caliber
AI capabilities, AI engines, AI models
AI capabilities: General character recognition, general form recognition, general multi-ticket recognition, general card recognition, text classification, information extraction and other AI capabilities
AI engine: Several versions are available for each capability, depending on the OCR provider or different processing mode. Universal character recognition, for example, can be standard or third-party. For example, text classification can choose to use the basic version of keyword features, but also can choose to support training intelligent version.
AI model: Multiple AI models can be created for each AI capability.
Each model has its own Pubkey and Secret.
The platform is limited to creating up to 200 AI models per capability.
The invocation of capabilities takes place in the AI model.
Statistical granularity - product scope
You can select different AI abilities, AI engines, AND AI models and count them together.
Example 1: You can choose model A for General table Recognition (standard edition) and model B for General word Recognition (standard edition).
Example 2: You can choose generic table recognition (standard version). At this point, all AI models representing general Word Recognition (standard version) are selected.
The user can only see and select AI capabilities within the account opening range.
Statistical granularity - time range
Today: Indicates the latest full hour from 00:00 to the current time. For example, the "Today" statistics queried at 13:30 p.m. are all the adjusted usage from 00:00 to 13:00 p.m.
This week: Last full natural day from Monday 00:00:00 to the current query time. For example, the query of This Week statistics on Thursday indicates all the query requests from 00:00.00 on Monday to 23:59:59 on Wednesday.
Current month: Indicates the latest full natural day from 00:00 on the first day of the current month to the current query time. For example, the statistics of "This Week" queried on 14th indicates the volume of all calls from 00:00:0 on 1st to 23:59:59 on 13th.
Custom time: Any time span that cannot exceed the current time. (The system only keeps 3 months of historical data)
Statistics - tune amount
The total number of successful calls to an account within a product range and time range.
The unit of adjustment amount is "times".
Failed calls are not counted in the call usage statistics.
The number of calls to all AI models that match a statistical caliber adds up to the number of calls for that caliber.
Example 1: Continuation of example 1 above, call times of AI model A is 10000 times, call times of AI model B is 10000 times, call amount is 20000 times.
Tuning amount visualization option - aggregation mode
Aggregation modes include Aggregation by AI capability, Aggregation by AI engine, and aggregation by AI model.
Continuation of example 2 above,
According to "aggregation by AI model", all AI models of the selected product range will be displayed in the consumption trend diagram, and each AI model corresponds to a curve.
According to "aggregation by AI Engine", all AI models of the selected product range will be displayed in the consumption trend diagram, and each AI engine to which the AI model belongs corresponds to a curve.
According to "aggregation by AI capability", all AI models of the selected product range will be displayed in the consumption trend chart, and each AI model belongs to a corresponding CURVE of AI capability.
Adjust volume visualization options - cumulative \ non-cumulative
Cumulative display data: The current day's adjustment data is the sum of all previous days' adjustment data.
Non-cumulative display data: The current day's adjustment data is the current day's adjustment data, excluding the previous adjustment data.
Assume that developer A, in the "official phase", has enabled all AI capabilities, namely general word recognition (standard) and text classification (Smart). He created two AI models based on general word recognition (standard version), namely application A and application B. Two AI models are created under text classification (intelligent version), which are application C and application D respectively. In the past two complete days, the dosage of the four AI models on the first day were 10000, 20000, 30000 and 40000 respectively. In the second day of the dosage is 10000, 20000, 30000, 40000.
Configuration item 1
Developer, check app A and App C in the product range. In the time range, the time range is defined as the last 2 days, and the display option is selected. "Aggregate by AI model" and "Non-cumulative display data" are selected.
Call usage = Call usage of Application A in the last two days + Call usage of application C in the last two days = 10000 + 30000 + 10000 + 30000 = 80,000.
On the trend chart, two curves representing application A and application C will be shown. The curve that applies A has two points, 10000 and 10000. The curve that applies C has two points, 30,000 and 30,000.
Configuration item 2
Developer, check app A and App C in the product range. In the time range, the time range is defined as the latest 2 days, display option, select "aggregate by AI model" and cumulative display data
On the trend chart, two curves representing application A and application C will be shown. The curve that applies A has two points, 10,000 and 20,000. The curve that applies C has two points, 30000 and 60000 respectively.
Configuration item 3
Developer, check app A, App B, app C in the product range. In the time range, a customized time range is set to the latest 2 days. In the display option, select "Aggregate by AI capability" and Cumulative display data
Call volume = Call volume of Application A in the last two days + Call volume of application B in the last two days + Call volume of Application C in the last two days = 10000 + 20000+ 30000 + 10000 + 20000+ 30000 = 100000.
On the trend chart, two curves will be shown, representing the general text recognition of AI abilities for apps A and B (standard edition) and the text classification of AI abilities for Apps C (Smart Edition). The curve for universal text recognition (standard edition) has two points, 30000 and 60000 respectively. For text classification (smart), the curve has two points, 30000 and 60000.
This is because, at this point, the calls of the two AI models under general text recognition (standard edition) will be added, which is what "aggregation" means.