DESC:FIELD OF INVENTION
[1] The embodiments herein generally relate to Knowledge Discovery and forecasting by Non-expert-user, using AI in multiple disconnected databases.
BACKGROUND OF INVENTION
[2] Databases are an integral part of modern organizations where all structured and unstructured data about the organization, operations, and business information is stored. Sometimes organizations have multiple databases for storing this data. These multiple databases could be due to a variety of reasons including the presence of some legacy database that could not be merged in the new database or very different data architectures (for example in a manufacturing company ERP database system may be used for their business and financial operations whereas SCADA database for their manufacturing operations) or you to differences in data basis of two different companies prior to merger.

[3] These databases are primarily relational databases having multiple tables with each table having multiple columns. Some columns are primary keys whereas some columns could be foreign keys related to the primary keys of other tables. Moreover, the column names and the tables at the time of design of such databases specific views which may be useful to the end user are created using SQL queries from the database. The results of queries are displayed either in tabulated or graphical form. Therefore, the end user is restricted to the possible views they can see on the database. For them to be able to see any other view they need to have a sufficient understanding of the SQL language and the underlying database architecture. Since the end users are not necessarily proficient in database knowledge this becomes a limitation in extracting the full use of the information present in the database. This problem gets further exaggerated if there are multiple databases that are not connected to each other.

[4] Therefore, in this invention, a method is described that provides access to different views of databases without the need for the end user to understand SQL queries or the underlying database architecture. Users can perform comprehensive database queries without needing expertise in SQL or knowledge of individual database structures by communicating with the system in natural human language. The invention further facilitates queries between multiple databases which are otherwise not connected. Finally, the embodiment of the invention provides a methodology for AI to choose and train machine learning algorithms for performing forecasting future data.

[5] The forecasting could be for any of the data available as a time series in the database which can include sales forecasting, demand forecasting, future pricing forecasting, machine breakdown forecasting, etc. This would help organizations make decisions based on factual data from the past and can perform future planning based on forecasting of specific events.
SUMMARY OF INVENTION
[6] The invention proposes a system and method for querying databases and performing machine learning-based forecasting using natural language inputs, thus enabling users to execute complex queries without prior knowledge of database architecture or query language.

[7] Database Querier:
Inputs from users in natural language are received via an input device (110) and processed by a Text Parser (120). The Text Parser converts user requests into SQL queries based on database architecture and forwards them to a Database Querier (130). The Database Querier (130) executes these queries using a Query Execution Module(131) on either a single database (132) or Multiple Databases (133). Results from the queries are transmitted to a Results Visualization Display (140) for presentation in various formats such as tables (141), graphs (142), or other visualizations modules (143).

[8] Forecaster:
Users provide input for forecasting via the input device (110), which is then parsed by the Text Parser (220). The Text Parser (220) identifies relevant data in the database and determines the time period for forecasting. Database queries are executed to gather data for training a machine learning model, performed by the Machine Learning Forecaster (230). The Forecaster aggregates data using Data Aggregator (231), selects appropriate machine learning models, trains these models, and then runs inference models for the specified time period. Forecast results are presented through the Results Visualization Display (140) in various formats.
[9] Advantages:
? Enables users to perform complex queries without prior knowledge of database architecture or query language.
? Facilitates querying across multiple databases not traditionally linked.
? Provides machine learning-based forecasting of database data.
? This is valuable for users seeking insights from databases without technical expertise.
? Useful for decision-making processes requiring data-driven forecasting.

[10] In conclusion, the invention presents an innovative approach to database querying and forecasting, leveraging natural language inputs and machine learning techniques, thereby enhancing accessibility and usability for users across diverse applications.

BRIEF DESCRIPTION OF DRAWINGS
[11] Embodiment of the invention will be now described using drawings. The drawings described here and are only one of the many possible implementations of the invention, and are not intended to limit the scope of the present disclosure.
[12] Figure 1 shows the block diagram of an embodiment of the present invention where the system performs the query and result visualization from various databases based on Dynamic input. This comprises of an input device (110), a text parser (120), a database querier (130), and a result visualization display (140).
[13] Figure 2 shows possible sub-components of systems described in Figure 1. The input device (110) may comprise of a microphone(111) and a text-to-speech engine(112); the text parser(120) may comprise of natural language processing engine(121), table & column identifier (122), and query identifier (123); the database querier (130) may comprise of query execution module (131) which may take data from either a single database(132) or multiple databases(133); the result visualization display (140) may comprise of result table (141), result graph (142) or other visualisation modules (143).
[14] Figure 3 comprises another possible in embodiment of the present invention as described in Figure 1. In this figure the input device (110) comprises of a keyboard (113); however, it could comprise of any other device that is capable of giving an input.
[15] Figure 4 shows the block diagram of an embodiment of the present invention where the system performs forecasting based on past data from the database. This may comprise of an input device (110), a text parser (220), a database querier (130), an ML forecaster (230), and a result visualization display (140).
[16] Figure 5 shows possible sub-components of the system described in Figure 4. The input device (110) may comprise of a microphone (111) and a text-to-speech engine (112); the text parser (220) may comprise of natural language processing engine (221), table & column identifier (222), and date stamp identification engine (223), forecasting problem identification engine (224); the database querier (130) may comprise of query execution module (131) which may take data from either a single database (132) or multiple databases (133); the ML forecaster (230) may comprise of Data Aggregator (231), ML model selector (232), which selects models from an ML model list (233), ML Model trainer (234), ML model inferences Module (235); the result visualization display (140) may comprise of result table (141), result graph (142) or other visualization modules (143).
[17] Figure 6 comprises of another possible embodiment of the present invention as described in Figure 4. In this figure the input device (110) may comprise of a keyboard (113); however, it could comprise of any other device that is capable of giving an input.

BRIEF DESCRIPTION OF INVENTION
[18] The invention takes inputs from an individual in natural language to perform a specific query in a database or multiple sets of databases (Figures 1, 2, and 3). The user input is taken in an input device (110) and passed to a natural language text parser(120). The Text Parser (120) comprehends which column of which table is being requested by the user and what query is being requested. This information is created into a SQL query based on the architecture of the database and given to a database querier (130). The database querier (130) uses this SQL Query to execute the database query using the Query Execution module (131) on either a single database (132) or multiple databases (133). The results of this query are then passed on to a results visualization display (140). The results visualization display (140) comprehends the data available in the results and accordingly presents the results either in the form of a table (141), a graph (142), or some other visualization modules (143). The input device could be any device that can provide the user with the intent to use the natural language text parsing algorithm. This may include a microphone (111) and a speech-to-text engine (112), a keyboard (113), or any other input means.
[19] Therefore, a user can perform Complex queries from a database using natural language, without knowing the database architecture or having any knowledge of the query language.
[20] This also provides a methodology to perform queries across multiple databases that are otherwise not linked together and cannot be queried by just a simple SQL statement.
[21] In another method of this invention, a machine learning-based forecasting of any database data can be performed (Figures 4, 5, and 6). In this system, again, there is an input device (110) that takes the user inputs on what quantity from the database the user wants to perform the forecasting on and the time over which forecasting has to be performed. This text is passed to a text parser (220), which comprehends the natural language and determines which table and which column the said data exists, looks at which possible quantities that vary with time (date stamp identification engine (223)), and provides input to the database querier (130) and an input to the machine learning forecaster (230). The database querier (130) performs the database query using query execution module (131) over a single database (132) or multiple databases (133) and creates a data set. This data set is used by the machine learning forecaster (230) to train the machine learning model. The machine learning forecaster(230) aggregates the data (231), chooses the possible machine learning models (232) from a list of models (233), performs model training (234) based on the data provided by the aggregator, and then runs inference module (235) for the time period as provided by the user as inferred by the text parser. The output of the forecasting is taken to the results visualization display (140), where the results of the forecast are either displayed as tables (141), graphs (142), or other modules (143). The input can be provided either by a combination of microphone-based audio input (111) and a speech-to-text engine (112), or by a keyboard (113), or by any other similar means.

DETAILED DESCRIPTION OF INVENTION

[22] This invention pertains to a system and method for querying databases and performing data forecasting through natural language input across multiple databases. The system employs AI/ML for parsing user queries, executing SQL commands across multiple databases, and utilizing machine learning techniques for forecasting database information.
[23] Input Device (110) Figure 1,2,3
[24] This takes user input in a natural language who has no prior knowledge of the database architecture or SQL query language. This accommodates various input devices, including microphone-based audio input with speech-to-text conversion, keyboards, or similar input mechanisms.
[25] This may comprise of the following
? Microphone (111) Figure 2
? Text to speech engine (112) Figure 2
[26] In another employment of this invention, this may comprise of
? Keyboard (113) Figure 3
? Or any other text input device
[27] Text Parser (120) Figure 1,2,3
[28] A natural language text parsing algorithm interprets user queries to identify specific database columns and tables and formulates corresponding SQL queries.
[29] This may comprise of the following
? NLP engine (121) Figure 2,3
? Table and column identifier (122) Figure 2,3
? Query identifier (123) Figure 2,3

[30] DB querier (130) Figure 1,2,3
[31] The invention facilitates querying databases based on user input without requiring knowledge of the underlying database architecture or query language. The SQL queries provided by Text Parser (120) are executed by a database querier (130) across single or multiple databases, producing query results.
[32] This may comprise of the following
? Query execution module (131) Figure 2,3
? Database (132) Figure 2,3
? Multiple databases (133) Figure 2,3
[33] Method for Cross-Database Queries
[34] This enables complex queries across multiple databases that are not inherently linked. This is achieved by the query execution module (131) which runs queries across multiple databases and joins them by unions and intersections to arrive at the unified result by quarrying across multiple databases.
[35] Results visualization display (140) Figure 1,2,3
[36] The results visualization display processes the query results and presents them in various formats, including tables, graphs, or other visualization modules.
[37] This may comprise of the following
? Result table (141) Figure 2,3
? Result graph (142) Figure 2,3
? Other visualization modules (143) Figure 2,3
[38] Users can perform comprehensive database queries without needing expertise in SQL or knowledge of individual database structures by communicating with the system in natural human language.
[39] Forecasting Methodology Figure 4,5,6
[40] The invention includes a machine learning-based forecasting method for database data. The users specify the data quantity and time period for forecasting through an input device.
[41] Input Device (110) Figure 4,5,6
[42] This takes user input in a natural language who has no prior knowledge of the database architecture or SQL query language. This accommodates various input devices, including microphone-based audio input with speech-to-text conversion, keyboards, or similar input mechanisms.
[43] This may comprise of the following
? Microphone (111) Figure 5
? Text-to-speech engine (112) Figure 5
[44] In another employment of this invention, this may comprise of
? Keyboard (113) Figure 6
? Or any other text input device
[45] Text Parser (220) Figure 4,5,6
[46] A natural language text parsing algorithm interprets what quantity from the database the user wants to perform the forecasting on and the time over which forecasting has to be performed. It identifies specific database columns and tables, looks at which possible quantities that vary with time, and provides input to the database query and input to the machine learning forecasting.
[47] This may comprise of the following
? NLP engine (221) Figure 5,6
? Table and column identifier (222) Figure 5,6
? Date stamp identification Engine (223) Figure 5,6
? Forecasting problem identification Engine (224) Figure 5,6
[48] DB Querier (130) Figure 4,5,6
[49] The invention facilitates querying databases based on user input without requiring knowledge of the underlying database architecture or query language. The SQL queries provided by Text Parser (220) are executed by a database querier across single or multiple databases, producing query results.
[50] This may comprise of the following
? Query execution module (131) Figure 5,6
? Database (132) Figure 5,6
? Multiple database (133) Figure 5,6
[51] Method for Cross-Database Queries
[52] This enables complex queries across multiple databases that are not inherently linked. This is achieved by the query execution module (131) which runs queries across multiple databases and joins them by unions and intersections to arrive at the unified result by quarrying across multiple databases.

[53] ML forecaster (230) Figure 4,5,6
[54] The data set received from the DB querier (130) is utilized to train machine learning models for forecasting. The ML Forecaster (230) selects appropriate models, trains them using the dataset, and generates forecasts for the specified time period.
[55] This may comprise of the following
? Data Aggregator (231)
? ML model selector (232)
? ML model list (233)
? ML Model trainer (234)
? ML model inferences module (235)
[56] Results visualization display (140) Figure 4,5,6
[57] Results of the forecasting process are displayed via the results visualization display in formats such as tables, graphs, etc.
[58] This may comprise of the following
? Result table (141) Figure5, 6
? Result graph (142) Figure 5, 6
? Other visualization modules (143) Figure 5,6
[59] This invention revolutionizes database querying and forecasting by enabling users to interact with databases using natural language and employing advanced algorithms for efficient data retrieval and forecasting across multiple databases.
,CLAIMS:I/We claim:
1. A system (100) for, non-expert-user driven, real-time, multiple disconnected databases queries, the system comprises of,
a. Input Device (110) for taking user input
b. Text Parser (120) for comprehending the user input and creating a DB query
c. DB Querier (130) for running the DB query in the DB, and
d. Result Visualization Display (140) for visualization of query results
2. The system (100) as claimed in claim 1, wherein the Input Device (110) comprises of at least one of keyboard (113), or a combination of Microphone (111) and Speech-to-Text-Engine (112).
3. The system (100) as claimed in claim 1, wherein the Text Parser (120) comprises of NLP Engine (121), Table & Column Identifier (122) and a Query Identifier (123)
4. The system (100) as claimed in claim 1, wherein the DB Querier (130) comprises of a Query Execution Module (131) and at least a single Database (132) or Multiple Databases (133).
5. The system (100) as claimed in claim 1, wherein the Result Visualization Display (140) comprises of at least one of Results Table (141), or Results Graph (142) or Other Visualization Module (143).
6. The system (200) which extends the claim 1 by performing non-expert-user driven data forecasting for the data in the database, the system comprised of,
a. Input Device (110) for taking user input
b. Text Parser (220) for comprehending the user input
c. DB Querier (130) for running the DB query in the DB
d. ML Forecaster (230) for running ML forecasting model, and
e. Result Visualization Display (140) for visualization of forecasting results
7. The system (200) as claimed in claim 6, wherein the Input Device (110) comprises of at least one of keyboard (113), or a combination of Microphone (111) and Speech-to-Text-Engine (112).
8. The system (200) as claimed in claim 6, wherein the Text Parse (220) comprises of NLP Engine (221), Table & Column Identifier (222) and a Date Stamp Identification Engine (223) and Forecasting Problem Identification Engine (224)
9. The system (200) as claimed in claim 6, wherein the DB Querier (130) comprises of a Query Execution Module (131) and at least a single Database (132) or Multiple Databases (133).
10. The system (200) as claimed in claim 6, wherein the ML Forecaster (230) comprises of Data Aggregator (231), ML Model Selector (232), ML Model List (233), ML Model Trainer (234) and ML Model inference Module (235).
11. The system (200) as claimed in claim 6, wherein the Result Visualization Display (140) comprises of at least one of Results Table (141), or Results Graph (142) or Other Visualization Module (143).
12. A method, for non-expert-user driven, real-time database queries, that runs on system (100), comprising steps of:
a. Receiving user input from Input Device (110) and passing it to a Text Parser (120)
b. The Text Parser (120) converts the text to a DB query and passes it to a DB querier (130)
c. The DB querier (130) executes the query.
d. The results of the DB querier (130) is passed to the Result Visualization Display (140)
13. The method of claim 12, wherein the Input Device (110) either receives user input from a Keyboard (113) or from a combination of Microphone (111) which receives a command over speech from the user and sends it to the Speech-to-Text-Engine (112) for conversion of speech to text.
14. The method of claim 12, wherein the
a. Text Parser (120) receives the text from the Input Device (110) and sends it to NLP Engine (121) for parsing.
b. The NLP Engine (121) send parsed text to the Table and Column Identifier (122) which identifies the DB table and column for which the user has requested in the input.
c. The output of NLP Engine (121) and the Table and Column Identifier (122) is then send to Query Identifier to create a SQL query on the DB table and column as requested by the user.
15. The method of claim 12, wherein the
a. DB Querier (130) receives a parsed DB query from Text Parser (120) and executes the query using Query Execution Module (131)
b. The Query Execution Module (131) runs the query on a single Databases (132) or can dynamically perform query on data distributed across Multiple Databases (133).
16. The method of claim 12, wherein the
a. Result Visualization Display (140) receives result of query from DB querier (130).
b. The Result Visualization Display (140), displays the results of the DB querier (130) on either a Results Table (141), or Results Graph (142) or Other Visualization Module (143).
17. A method, for non-expert-user driven data forecasting for the data in the database, that runs on system (200), comprising steps of:
a. Receiving user input from Input Device (110) and passing it to a Text Parser (220)
b. The Text Parser (220) converts the text to a DB forecasting query and passes it to a DB querier (130)
c. The DB querier (130) executes the query to create training data.
d. The results of the DB querier (130) is passed to ML Forecaster (230) which runs ML to perform forecasting
e. The ML prediction of the ML Forecaster (230) is passed to the Result Visualization Display (140) for visualization
18. The method of claim 17, wherein the Input Device (110) either receives user input from a Keyboard (113) or from a combination of Microphone (111) which receives a command over speech from the user and sends it to the Speech-to-Text-Engine (112) for conversion of speech to text.
19. The method of claim 17, wherein the
a. Text Parser (220) receives the text from the Input Device (110) and sends it to NLP Engine (221) for parsing.
b. The NLP Engine (221) send parsed text to the Table & Column Identifier (222) which identifies the DB table and column for which the user has requested the forecast.
c. The input form NLP Engine (221) and the Table & Column Identifier (222) is send to the Date Stamp Identification Engine (223), which identifies the future date for which the forecast has been requested.
d. The forecasting time from the Date Stamp Identification Engine (223) and the DB table and column from the Table & Column Identifier (222) are received by Forecasting Problem Identification Engine (224).
e. The Forecasting Problem Identification Engine (224) creates the DB query for generating test data and ML problem statement. The DB query for generating test data is send to the DB Querier (130) and the ML problem statement is send to ML Forecaster(230).
f. The result generated by DB querier (130) is send to ML Forecaster (230).
g. The ML Forecaster (230) receives the ML problem statement from the Forecasting Problem Identification Engine (224) and trains a ML model based on the result generated by DB querier (130) and performs the forecasting.
h. The ML Forecaster (230) sends the ML forecasting result to the Result Visualization Display (140) for forecasting visualization.
20. The method of claim 17, wherein the ML Forecaster (230) comprising steps of:
a. Aggregating the results in Data Aggregator (231) received from DB Querier (130) to create a training data set.
b. Passing the ML problem statement from the Forecasting Problem Identification Engine (224) of the Text Parser (220) to the ML Model Selector (232)
c. The ML Model selector (232) takes the ML problem statement and selects the relevant ML model from the ML Model List (233)
d. The selected ML Model and the Aggregated results are send by ML Model Selector (232) to the ML Model Trainer (234)
e. The ML Model Trainer (234) performs training of the ML model based on the Aggregated results and sends the trained model to the ML Model Inference Module (235).
f. The ML Model Inference Module (235) runs the trained ML model for the dates provided by Date Stamp Identification Engine (223) of the Text Parser (220).
g. The results of the ML Model Inference Module (235) are passed to Result Visualization Display (140)
21. The method of claim 17, wherein the
a. Result Visualization Display (140) receives forecasting result from ML Model Inference Module (235) of the ML Forecaster (230)
b. The Result Visualization Display (140), displays the forecasting results on either a Results Table (141), or Results Graph (142) or Other Visualization Module (143).