CN111090707B - Data processing method and device, electronic equipment and storage medium - Google Patents
Data processing method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111090707B CN111090707B CN201811244716.0A CN201811244716A CN111090707B CN 111090707 B CN111090707 B CN 111090707B CN 201811244716 A CN201811244716 A CN 201811244716A CN 111090707 B CN111090707 B CN 111090707B
- Authority
- CN
- China
- Prior art keywords
- classification
- data
- model
- sample data
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 18
- 238000003860 storage Methods 0.000 title claims abstract description 16
- 238000013145 classification model Methods 0.000 claims abstract description 129
- 238000012544 monitoring process Methods 0.000 claims abstract description 115
- 238000013507 mapping Methods 0.000 claims abstract description 47
- 238000000034 method Methods 0.000 claims abstract description 41
- 238000002372 labelling Methods 0.000 claims description 57
- 238000012545 processing Methods 0.000 claims description 8
- 230000000875 corresponding effect Effects 0.000 description 19
- 239000002689 soil Substances 0.000 description 18
- 238000012549 training Methods 0.000 description 16
- 238000010586 diagram Methods 0.000 description 8
- 238000003973 irrigation Methods 0.000 description 8
- 230000002262 irrigation Effects 0.000 description 8
- 241000238631 Hexapoda Species 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 230000001960 triggered effect Effects 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 4
- 238000013550 semantic technology Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 241000607479 Yersinia pestis Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000003631 expected effect Effects 0.000 description 1
- 230000004720 fertilization Effects 0.000 description 1
- 230000000749 insecticidal effect Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000000575 pesticide Substances 0.000 description 1
- 235000003784 poor nutrition Nutrition 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 238000005507 spraying Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a data processing method and device, electronic equipment and a storage medium. The method further comprises the following steps: based on a classification model, obtaining first clustering information of first sample data, and determining a selection threshold according to the first clustering information; acquiring second clustering information of first monitoring data of a target based on the classification model, and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold; performing active learning of the classification model based on the second sample data, and optimizing model parameters of the classification model; classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data; and mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and obtaining an event decision for executing a preset operation on the target based on the updated body.
Description
Technical Field
The present invention relates to the field of information technologies, and in particular, to a data processing method and apparatus, an electronic device, and a storage medium.
Background
The learning model can obtain model parameters through the learning of sample data; so as to make various decisions by using the trained model; for example, application to agricultural systems to provide plant-related decisions.
In the process of training the model, the decision quality of the model is directly determined by the amount of sample data and the comprehensiveness of values. It is almost impossible to obtain data in all cases. For example, taking a learning model applied to an agricultural system as an example, various conditions such as weather and crops can be dynamically changed, and sample data is difficult to collect completely, so that the accuracy of decision obtained by the learning model cannot achieve the expected effect.
Disclosure of Invention
In view of this, embodiments of the present invention are intended to provide a data processing method and apparatus, an electronic device, and a storage medium.
The technical scheme of the invention is realized as follows:
a method of data processing, comprising:
acquiring first clustering information of first sample data based on a classification model, and determining a selection threshold according to the first clustering information;
acquiring second clustering information of first monitoring data of a target based on the classification model, and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold;
performing active learning of the classification model based on the second sample data, and optimizing model parameters of the classification model;
classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
and mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and obtaining an event decision for executing a preset operation on the target based on the updated body.
Based on the above scheme, the obtaining first clustering information of the first sample data based on the classification model includes:
based on the classification model, obtaining the clustering average distance of the first sample data, the maximum distance between the sample and a clustering center and the classification accuracy;
the determining a selection threshold according to the first clustering information includes:
and calculating the selection threshold according to the clustering average distance, the maximum distance between the sample and the clustering center and the classification accuracy.
Based on the above scheme, the obtaining of the second classification information of the first monitoring data based on the classification model includes:
clustering first monitoring data based on the classification model to obtain a clustering distance between the first monitoring data and a corresponding clustering center;
selecting second sample data from the first monitoring data according to the second classification information and the selection threshold, wherein the selecting of the second sample data comprises:
and selecting a part of the first monitoring data with the clustering distance larger than the selection threshold value as the second sample data.
Based on the above solution, the second aggregation information further includes: obtaining a classification label of the second sample data by using the classification model;
the method further comprises the following steps: acquiring a manual labeling label manually labeled by the second sample data;
and if the classification label of the second sample data is different from the artificial labeling label, actively learning the second sample data and the artificial labeling label by the classification model so as to optimize the model parameter of the classification model.
Based on the above scheme, the method further comprises:
updating the classification accuracy of the classification model according to the comparison result of the classification label of the second sample data and the manual labeling label;
obtaining third classification information of the first sample data and the second sample data by using the updated classification model;
and updating the selection threshold according to the updated classification accuracy and the third classification information.
Based on the above scheme, the method further comprises:
if the classification label of the second sample data is different from the manual labeling label, performing semantic mapping on the manual labeling label to obtain semantic information of the manual label; updating the ontology knowledge model based on semantic information of the artificial tags, and performing event decision of executing a predetermined operation on the target by the updated ontology knowledge model;
or,
if the classification label of the second sample data is the same as the manual labeling label, performing semantic mapping on the manual labeling label or the classification label of the second sample data to obtain semantic information of the second sample data; and updating the ontology knowledge model based on the semantic information of the second sample data, and performing event decision of executing a preset operation on the target by the updated ontology knowledge model.
Based on the above scheme, the method further comprises:
if the first monitoring data is not selected as the second sample data, at least performing semantic mapping on the classification label in the second classification information to obtain semantic information of the first monitoring data;
and updating the ontology knowledge model based on the semantic information of the first monitoring data, and performing event decision of executing a preset operation on the target by the updated ontology knowledge model.
A data processing apparatus comprising:
the threshold value determining module is used for acquiring first clustering information of the first sample data based on the classification model and determining a selection threshold value according to the first clustering information;
the selection module is used for acquiring second clustering information of the first monitoring data of the target based on the classification model and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold;
the active learning module is used for actively learning the classification model based on the second sample data and optimizing the model parameters of the classification model;
the classification module is used for classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
and the mapping decision module is used for mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and acquiring an event decision for executing a preset operation on the target based on the updated body.
An electronic device, comprising:
a memory for information storage;
and the processor is connected with the memory and is used for realizing the data processing method provided by one or more of the technical schemes by executing the computer executable instructions stored on the memory.
A computer storage medium having computer executable code stored thereon; the computer executable code can implement the data processing method provided by one or more of the above technical solutions after being executed by a processor.
According to the data processing method provided by the embodiment of the invention, first clustering information is obtained according to first sample data, and then second clustering information is classified on the first monitoring data by using a current classification model after the first monitoring data needing to be classified is obtained; selecting second sample data needing active learning from the first monitoring data by combining the second clustering information and a selection threshold value obtained based on the first clustering information; therefore, the classification model also performs active learning in the application process after passive learning, so that the training data of the classification model is more comprehensive, and the phenomenon of inaccurate classification caused by incomplete or insufficient abundant sample data is reduced; and the training can be carried out while applying, so that the data actually used for classification in the application process participates in the training of the classification model, and the method has the characteristics of convenience and simplicity in obtaining sample data of model training. Furthermore, to enable automatic event decision-making; the ontology knowledge model is introduced to realize the event decision, but the ontology knowledge model generally can only not identify the numerical classification labels, so in this embodiment, the optimized classification model is used for classifying the second monitoring data, the obtained classification labels are mapped into semantic information, the ontology is updated by using the semantic information, and the updated ontology knowledge model automatically carries out event decision based on the updated ontology; and automation of event decision is realized.
Drawings
Fig. 1 is a schematic flow chart of a data processing method according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a cluster according to an embodiment of the present invention;
fig. 3 is a schematic flow chart of a data processing method according to an embodiment of the present invention;
FIG. 4 is a diagram of an ontology repository provided by an embodiment of the present invention;
fig. 5 is a schematic flowchart of a data processing method according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a data processing apparatus according to an embodiment of the present invention;
FIG. 7 is a block diagram of a data processing system in accordance with an embodiment of the present invention;
fig. 8 is a flowchart illustrating a data processing method according to an embodiment of the present invention;
fig. 9 is a schematic flowchart of a data processing method according to an embodiment of the present invention.
Detailed Description
The technical solution of the present invention is further described in detail with reference to the drawings and the specific embodiments of the specification.
As shown in fig. 1, the present embodiment provides a data processing method, including:
step S110: based on a classification model, obtaining first clustering information of first sample data, and determining a selection threshold according to the first clustering information;
step S120: acquiring second clustering information of first monitoring data of a target based on the classification model, and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold;
step S130: performing active learning of the classification model based on the second sample data, and optimizing model parameters of the classification model;
step S140: classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
step S150: and mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and obtaining an event decision for executing a preset operation on the target based on the updated body.
The classification model can be various models with learning ability trained by sample data, such as deep learning models like neural networks, or machine learning models like vector machines. The classification model may correspond to various classification algorithms, such as distance-based classification, density-based classification, etc.; such as a binary tree or a multi-tree classification algorithm.
After the electronic equipment inputs the first monitoring data into the classification model by operating the model, the classification model can automatically output the second classification information. For example, the second classification information may include at least: and (5) classifying the labels. For another example, the second classification information may include, in addition to the classification tag: classification confidence or distance between the first monitored data and the center point of the class to which the first monitored data is classified. Of course, this is merely an example of the first classification information, and the specific implementation is not limited thereto.
The active learning is as follows: and under the condition of no manual trigger of the user, the equipment trains the classification model based on the self-determined second sample data, so that the model parameters of the classification model are updated. For example, if the classification model is a neural network, updating the model parameters may include: and updating the weight value and/or the threshold value of the neural network.
In this embodiment, in a first aspect, after the classification model obtains the first monitoring data, it is determined whether active learning needs to be started according to the second classification information obtained in the process of classifying the first monitoring data. Therefore, the classification model is used for providing clustering information needed by decision-making in a classification mode, and meanwhile active learning can be achieved in the using process, so that the classification model can obtain training of more samples, and the classification result of the classification model is more accurate. In a second aspect, not all the first monitoring data can trigger the classification model to perform active learning, but the currently obtained second classification information and the selection threshold are judged, and second sample data capable of improving the classification capability of the classification model is selected, so that unnecessary active learning is reduced. In a third aspect, in this embodiment, the selection threshold is determined according to the first clustering information, and the selection threshold represents a clustering result between sample data that has been used for training a classification model, so that the currently classifiable classification model represented by the selection threshold may be the range and the strength of data. If the difference between the first monitoring data and the first sample data is large, the first monitoring data and the selection threshold naturally meet a certain condition; the selection threshold value is dynamically determined by the threshold value device instead of a preset static value, and the selection threshold value fully reflects the current classification capability of the classification model, so that the active learning of the automatic triggering classification model is realized, and the active learning of the classification model can be triggered at the required time, thereby reducing the unnecessary learning.
In this embodiment, after the active learning of the classification model, the model parameters of the classification model are optimized, for example, taking the classification model as an example of a neural network, the network parameters of the neural network are optimized by the active learning, and the network parameters may include but are not limited to: the weight of each node in the network, etc.; the optimized classification model can be used for more accurately classifying the monitoring data of the target.
In the embodiment of the present invention, in step S120, the second sample data may be selected from the first monitoring data according to a preset time interval, for example, the second sample data is periodically selected from the first monitoring data. At this time, if the second sample data does not need to be selected, semantic mapping can be directly performed based on the classification label in the second classification information, the ontology is updated according to the semantic information obtained by the semantic mapping, and event decision is performed by the ontology knowledge model based on the updated ontology.
Classifying the second monitoring data by using the optimized classification model to obtain a classification label and the like; the classification label can be a label which can be recognized by various classification models which are digitalized or serialized. But to enable automatic execution of event decisions. In step S150 of this embodiment, the classification label is first mapped to semantic information that can be recognized by the ontology knowledge model, and the ontology in the ontology knowledge model is updated by using the speech information. The ontology knowledge model comprises: an ontology knowledge base and an ontology; the ontology knowledge base comprises entities and incidence relations between the entities, the ontology can record the current state or attribute value information of the target, and the ontology in the ontology knowledge model can be considered to be updated in step S150; and based on the updated ontology, making an event decision for executing a predetermined operation on the target according to the ontology knowledge base. And if the target is the crop, determining whether to carry out watering, pest killing and other operations on the crop.
The step S110 may include: and obtaining the clustering average distance of the first sample data, the maximum distance between the sample and a clustering center and the classification accuracy rate based on the classification model.
For example, the classification model may classify data into M classes, where the clustering centers of different classes of sample data are different, and the distance between the sample data of one class and the clustering center is also different. In this embodiment, the average distance of the clusters, the maximum distance between the sample of each class and the cluster center, and the classification accuracy rate can be obtained by M classes; then M selection thresholds are obtained. And if the second clustering information shows that the current first monitoring data belongs to the mth class, selecting a selection threshold value of the mth class to judge whether the corresponding first monitoring data needs to be used as second sample data for the active learning of the classification model. M is a positive integer not greater than M. For example, taking the classification of soil humidity as an example, there are three categories, namely "wet", "moderate" and "dry", then the three categories correspond to three selection thresholds, and if the current first monitoring data is determined to correspond to "wet" through the processing of the classification model, then the selection threshold corresponding to "wet" is selected to determine whether the corresponding first monitoring data constitutes the second sample data.
The step S120 may include: and calculating the selection threshold according to the clustering average distance, the maximum distance between the sample and the clustering center and the classification accuracy.
Center point c of ith sample i ，
i Represents a classification category, X j Is the jth sample in the i class, i having n samples in total, c i Is the cluster center of the ith category. Then all samples in each class are calculated to the cluster center c i Of (d) is the mean value of the Euclidean distances i (i.e., the cluster mean distance).
For example,
calculate all samples within each class to cluster center c i The maximum value D of the distance of (i.e. the maximum distance of the sample from the cluster center),
let the sample selection Threshold (Threshold) have an initial value of Threshold = d i +(D-d i ) X.p. p is the classification accuracy; (X) j -c i ) T Is (X) j -c i ) The transposing of (1).
Fig. 2 is a schematic diagram of a cluster, and fig. 2 shows a distribution of sample data of a category corresponding to a classification label.
The above is a way to determine the selection threshold, and the specific implementation is not limited to the above scheme; for example, in some embodiments, it may be directly determined whether the corresponding first monitoring data is used as the second sample data for the classification model to actively learn according to the maximum distance between the sample and the cluster center. For example, if the distance between the first monitoring data and the clustering center is greater than the maximum distance between the sample and the clustering center, the information to be classified may be selected as the second sample data, otherwise, the information is not selected as the sample data.
However, in this embodiment, the selection threshold is determined in combination with the cluster average distance, the maximum distance between the sample and the cluster center, and the classification accuracy, and the selection threshold thus determined may be used to accurately select whether a second sample data of a training classification model needs to be improved, so as to improve the classification capability of the classification model trained based on the second sample data.
In some embodiments, the step S120 may further include:
clustering first monitoring data based on the classification model to obtain a clustering distance between the first monitoring data and a corresponding clustering center;
and selecting a part of the first monitoring data with the clustering distance larger than the selection threshold value as the second sample data.
In this embodiment, if the clustering distance calculated in the first monitoring data is greater than the selection threshold, it indicates that a new class or a critical value of a certain class may appear at present, and autonomous learning of the classification model is required, and the first monitoring data is selected as second sample data for updating the model parameters of the classification model. In this embodiment, if the classification model is a neural network, the updating of the model parameters may include: updating the weight and/or threshold of the neural network, and the like.
For example, a certain first monitoring data a is attributed to the kth class by the classification model, and if the clustering distance between the first monitoring data a and the clustering center of the kth class is D1 and is greater than the selection threshold of the kth class, the first monitoring data a is selected as the second sample data to train the classification model, so that the clustering capability of the classification model on the data with the classification label of the kth class is improved. For another example, if the clustering threshold of the first monitoring data a is smaller than the selection threshold of the kth class, the first monitoring data a is not used as the second sample data for the active learning of the classification model because the value of the first monitoring data a for the active learning of the classification model is not large.
In some embodiments, the second classification information further comprises: obtaining a classification label of second sample data of the second sample data by using the classification model; as shown in fig. 3, the method further comprises:
step S160: taking a manual labeling label manually labeled by the second sample data;
step S170: if the classification label of the second sample data is different from the artificial labeling label, the classification model actively learns the second sample data and the artificial labeling label so as to optimize the model parameters of the classification model.
If the clustering threshold of the first monitoring data is greater than the corresponding selection threshold, it indicates that the accuracy of the classification label of the second sample data allocated to the first monitoring data by the current classification model may not be high. At this time, a corresponding professional person is required to perform manual labeling.
The active learning may include the following:
based on active learning of data flow, second sample data is submitted to a training engine of the classification model according to the sequence; the training engine trains the classification model according to the sequence determined by the second sample data;
based on the active learning of the data pool, after second sample data is obtained, the active learning of the classification model is not started directly, but enough second sample data is required to be accumulated, enough artifact data form the data pool, and the training engine trains the classification engine by using the second sample data in the whole data pool.
In this embodiment, the active learning may be any one of the above.
In some embodiments, the step S160 may include: the manual annotation tag is received from a human-machine interaction interface (e.g., keyboard, mouse, or voice interaction device).
In other embodiments, the step S160 may further include: a manually labeled manual label is received from the other device.
In this embodiment, before obtaining the manual labeling tag, the method may further include:
outputting the second sample data for manual marking;
or, outputting a labeling prompt, so that the electronic equipment can receive a manual labeling label manually labeled based on the labeling prompt.
If the manually labeled manual labeling label is the same as the classification label of the second sample data distributed by the classification model, the classification of the first monitoring data by the current classification model is accurate, the current classification model can accurately classify the first monitoring data, and the classification model is not updated and trained at the moment, namely, the classification model is not required to be actively learned based on the second sample data.
If the manual labeling label is different from the classification label of the second sample data, the classification model does not have the capability of accurately classifying the first monitoring data, so the classification model needs to be updated based on the manual labeling label of the manual labeling and the first monitoring data; namely, the classification model carries out active learning based on the second sample data and the artificial labeling label.
In some embodiments, the method further comprises:
updating the classification accuracy of the classification model according to the comparison result of the classification label of the second sample data and the manual labeling label;
obtaining third classification information of the first sample data and the second sample data by using the updated classification model;
and updating the selection threshold according to the updated classification accuracy and the third classification information.
If the classification label of the second sample data is different from the manual labeling label, the current classification accuracy is reduced, and if the classification label of the second sample data is the same as the manual labeling label, the classification model is correctly classified once again, so that the classification accuracy is possibly improved. Therefore, in this embodiment, the comparison result of the two classification tags is used to update the classification accuracy of the selection threshold calculated later.
If the second sample data is selected from the first monitoring data, the sample database of the classification model is updated, so that third classification information needs to be obtained based on the updated sample data to further update the selection threshold.
In some embodiments, the method further comprises:
if the classification label of the second sample data is different from the manual labeling label, updating the current recorded target state of the ontology knowledge model according to the manual labeling label; the ontology knowledge model is a decision model for making an event decision based on the classification labels of the classification model;
and if the classification label of the second sample data is the same as the manual labeling label, updating the current recorded target state of the ontology knowledge model according to the classification label or the manual labeling label of the second sample data.
For example, if the first monitoring data is the state of the farmland being monitored, the target state is the state of the farmland. For another example, if the first monitoring data is the status of the monitored crop, the target status is the status of the crop. For example, the status of the crop may include: the health state, which can be classified as: "healthy", "sub-healthy" or "diseased". And the like. Of course, the above is merely exemplary, and the specific implementation is not limited to the following.
In this embodiment, the system for deciding whether to execute an event further includes an ontology model, and an ontology knowledge base may be disposed in the ontology model. Storing correlated entities and concepts in the ontology knowledge base; the ontology knowledge base can perform decision on whether to execute an event and/or which event to execute based on the entity and the concept in the ontology knowledge base and on the association relationship between different entities in the ontology knowledge base when the classification tag corresponds to a certain entity.
If new second sample data is introduced, new entities and/or concepts may need to be extracted for the ontology knowledge base, so in this embodiment, the second sample data with the clustering distance greater than the selection threshold in the first monitoring data is used to update the ontology knowledge base, thereby implementing active learning of the ontology knowledge base and improving the decision quality of the ontology knowledge base.
FIG. 4 is a diagram of an ontology base corresponding to an ontology model, such as an agricultural system; the specific implementation is not limited to the ontology repository shown in fig. 3.
In some embodiments, the method further comprises:
if the classification label of the second sample data is different from the manual labeling label, performing semantic mapping on the manual labeling label to obtain semantic information of the manual label; based on semantic information of the artificial tag, making an event decision by the ontology knowledge model to perform a predetermined operation on the target;
or,
if the classification label of the second sample data is the same as the manual labeling label, performing semantic mapping on the manual labeling label or the classification label of the second sample data to obtain semantic information of the second sample data; and updating the ontology based on the semantic information of the second sample data, and performing event decision of executing a preset operation on the target by the ontology knowledge model based on the updated ontology.
The classification labels output by the classification model are numerical values, for example, numerical values such as "0", "1", and "2", which have no correspondence with the entities in the ontology knowledge model.
In this embodiment, in order to enable the ontology model to recognize the classification tags, the classification tags are semantically converted, for example, the digitized classification tags are converted into textual classification tags. For example, for soil moisture, the numerical values "0", "1", and "2" are converted to: the semantic information of the first monitoring data such as 'soil moisture', 'soil humidity is moderate', and 'soil drought'. In this way, the ontology knowledge model can recognize the semantic information of the first monitoring data, so as to determine whether or not to execute an event or which event to execute based on the semantic information of the first monitoring data corresponding to the classification tag.
The method provided by the embodiment can be applied to an agricultural system, but is not limited to the agricultural system, and for example, the method can also be used for an intelligent home control system and the like. For example, control of indoor humidity in smart home systems, and the like.
If the classification label is given to the currently acquired first monitoring data, the corresponding event needs to be triggered and executed in time according to the accurate classification label. Therefore, when the classification label of the second sample data is different from the artificial labeling label, the artificial labeling label is subjected to semantic conversion (namely semantic mapping) to form semantic information, and the event decision of executing the preset operation on the target is carried out by the ontology knowledge model.
If the classification label and the manual labeling label of the second sample data are available, one of the classification label and the manual labeling label of the second sample data can be selected at will for semantic mapping, and an event decision of executing a preset operation on the target is triggered by the ontology knowledge model.
In some embodiments, the method further comprises: if the first monitoring data is not selected as the second sample data, performing semantic mapping on at least the classification label in the second classification information to obtain semantic information of the first monitoring data;
and based on the semantic information of the first monitoring data, making an event decision for executing a predetermined operation on the target by the ontology knowledge model.
Fig. 5 is a schematic flowchart of a data processing method according to an embodiment of the present invention, which includes:
acquiring sensor data, wherein the sensor data is one of the first monitoring data;
the classification model (e.g., a sudden humidity model as shown in fig. 5) performs machine learning, performs classification, and obtains a classification label for the currently acquired sensor data. The classification label may be a numerical label. Through semantic mapping, semantic information that the ontology knowledge model can recognize, for example, "wet", "drought", and "moderate" as shown in fig. 5, can be mapped. The ontology knowledge model updates the currently recorded sudden humidity state, then queries an ontology knowledge base based on the currently updated humidity state, and makes an event decision for executing a predetermined operation on the target, for example, an irrigation event decision, based on a corresponding relationship between the sudden humidity state and the event in the ontology knowledge base.
The first monitoring data includes: farmland monitoring data;
wherein the farmland monitoring data comprises at least one of:
air temperature data of the position of the farmland;
air humidity data of the location of the farmland;
soil moisture data of the field;
soil temperature data of the farmland;
illumination data of the position of the farmland;
rainfall prediction data of the position of the farmland;
crop growth status data in the field;
crop health data in the field.
Correspondingly, the event comprises at least one of:
an irrigation event;
an insecticidal event;
a fertilization event;
a lighting adjustment event.
For example, the classification model outputs classification labels with insects if the crops are found to have insects through the classification of the first monitoring data, the numerical classification labels in the classification models are converted into semantic information of the first monitoring data of the crops with insects through semantic mapping, and the ontology knowledge model makes an insect killing event according to entities and concepts stored in the ontology knowledge model.
After the ontology model makes the event of disinsectization, the method further comprises:
outputting deinsectization prompt information to prompt a user to deinsectize crops;
according to the insect killing event, the intelligent device is triggered to automatically kill insects, for example, the automatic spraying device in the farmland is triggered to spray liquid containing pesticides, so that the insect killing is realized.
In summary, the method in this embodiment may further include:
outputting a prompt corresponding to the event according to the event determined by the ontology knowledge model;
or, directly controlling the corresponding intelligent equipment to execute the corresponding event according to the event, for example, triggering intelligent irrigation equipment to intelligently irrigate when the farmland is dry; when the crop is in poor nutrition, the intelligent fertilizing equipment is triggered to automatically fertilize, and the like.
As shown in fig. 6, the present embodiment provides a data processing apparatus including:
a threshold determining module 110, configured to obtain first clustering information of the first sample data based on the classification model, and determine a selection threshold according to the first clustering information;
a selecting module 120, configured to obtain second classification information of the target first monitoring data based on the classification model, and select second sample data from the first monitoring data according to the second classification information and the selection threshold;
an active learning module 130, configured to perform active learning of the classification model based on the second sample data, and optimize model parameters of the classification model;
the classification module 140 is configured to classify the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
and a mapping decision module 150, configured to map the classification label of the second monitored data into semantic information, update the ontology by using the semantic information, and obtain an event decision for executing a predetermined operation on the target based on the updated ontology.
In some embodiments, the threshold determination module 110, the selection module 120, the active learning module 130, the classification module 140 and the mapping decision module 150 may be program modules, which when executed by a processor enable the determination of the first clustering information, the second clustering information, the selection threshold and the second sample data.
In other embodiments, the threshold determination module 110, the selection module 120, the active learning module 130, the classification module 140, and the mapping decision module 150 may be a soft-hard combining module, such as a field programmable array or a complex programmable array.
In still other embodiments, the threshold determination module 110, selection module 120, active learning module 130, classification module 140, and mapping decision module 150 may be purely hardware modules, such as application specific integrated circuits.
In some embodiments, the threshold determining module 110 is specifically configured to obtain a clustering average distance of the first sample data, a maximum distance between a sample and a clustering center, and a classification accuracy based on the classification model, and calculate the selection threshold according to the clustering average distance, the maximum distance between a sample and a clustering center, and the classification accuracy.
In some embodiments, the selecting module 120 is specifically configured to cluster the first monitoring data based on the classification model, and obtain a clustering distance between the first monitoring data and a corresponding clustering center; and selecting the part of the first monitoring data with the clustering distance larger than the selection threshold value as the second sample data.
In some embodiments, the second classification information further comprises: and obtaining a classification label of second sample data of the second sample data by using the classification model. The device further comprises: the third obtaining module is specifically configured to obtain an artificial labeling label for the artificial labeling of the second sample data; and if the classification label of the second sample data is different from the artificial labeling label, actively learning the second sample data and the artificial labeling label by the classification model so as to optimize the model parameter of the classification model.
In still other embodiments, the apparatus further comprises:
the first updating module is specifically used for updating the classification accuracy of the classification model according to the comparison result of the classification label of the second sample data and the manual labeling label;
a fourth obtaining module, configured to obtain third classification information of the first sample data and the second sample data by using the updated classification model;
and the second updating module is specifically configured to update the selection threshold according to the updated classification accuracy and the third classification information.
In some embodiments, the apparatus further comprises:
the third updating module is used for updating the current recorded target state of the ontology knowledge model according to the manual labeling label if the classification label of the second sample data is different from the manual labeling label;
and the fourth updating module is used for updating the current recorded target state of the ontology knowledge model according to the classification label of the second sample data or the artificial labeling label if the classification label of the second sample data is the same as the artificial labeling label.
In some embodiments, the mapping decision module 150 is configured to perform semantic mapping on the manual tagging tag to obtain semantic information of the manual tagging if the classification tag of the second sample data is different from the manual tagging tag;
and the decision module is used for carrying out event decision on executing a preset operation on the target by the ontology knowledge model based on the semantic information of the artificial tag.
In other embodiments, the mapping decision module 150 is further configured to perform semantic mapping on the manual tagging label or the classification label of the second sample data to obtain semantic information of the second sample data if the classification label of the second sample data is the same as the manual tagging label; and updating the body based on the semantic information of the second sample data, and performing event decision for executing a predetermined operation on the target based on the updated body.
In some embodiments, the mapping decision module 150 is specifically configured to perform semantic mapping on at least a classification label in the second classification information to obtain semantic information of the first monitoring data, if the first monitoring data is not selected as the second sample data; and updating an ontology based on the semantic information of the first monitoring data, and performing event decision of executing a predetermined operation on the target based on the updated ontology.
In some embodiments, the first monitoring data comprises: farmland monitoring data;
wherein the farmland monitoring data comprises at least one of:
air temperature data of the position of the farmland;
air humidity data of the position of the farmland;
soil moisture data of the field;
soil temperature data of the farmland;
illumination data of the position of the farmland;
rainfall prediction data of the position of the farmland;
crop growth status data in the field;
crop health data in the field.
Several examples are provided below in connection with any of the above embodiments:
example 1:
the example provides an intelligent agriculture system based on active learning and semantic technology, and a system diagram is shown in FIG. 7. The system mainly comprises a passive learning module, an active learning module and a semantic knowledge base module.
The passive learning module obtains the passively learned farmland model based on historical farmland data through model training before the farmland model is applied online.
And the semantic knowledge base can be used for mapping the classification labels of the farmland model into semantic information which can be identified by the agricultural knowledge base. The agricultural knowledge base is one of the above ontology knowledge bases.
The ontology update here may be: updating the current farmland state, wherein the farmland state is one of the target states.
The active learning module is mainly used for actively learning the farmland model so as to update the farmland model.
The three modules are associated through mechanisms such as model updating, ontology updating and reasoning of active learning, event feedback and the like to form a set of agricultural Internet of things automatic control system with self-adaptive capacity. In agriculture, irrigation operation is an important operation, and the following scheme is implemented by taking irrigation operation as an example, and the specific process is as follows:
a passive learning module for learning a farmland model;
in the passive learning module, model training is carried out on farmland data marked based on history, and a farmland model is obtained through training, wherein the model represents farmland environment information of the past year or a past period of time.
Taking irrigation as an example, the historical farmland data here includes air temperature data, air humidity data, soil temperature data, soil humidity data, illuminance data, date data, rainfall data for two days in the future. The seven-dimensional data is used as a training sample for passive learning, a model related to farmland humidity is trained in a certain machine learning mode, and the trained model can divide farmland data into three types: "humid", "drought", "moderate". The 'humidity', 'drought' and 'moderate' are just classification labels, have the same labeling effect as the labeling effect of the '0', '1' and '2', have no semantics, and cannot be understood and shared by the equipment of the Internet of things. In the passive learning, in order to correspond to the subsequent model updating part, a softmax regression method is selected here for model training.
The semantic knowledge base module is used for performing semantic mapping, ontology updating and event decision;
in the agricultural internet of things, a plurality of interrelated entities and concepts exist, wherein the interrelated entities comprise agricultural execution equipment, agricultural sensors, crops, farmlands and the like, and the interrelated entities, the supporting and the cooperation of the interrelated entities and the agricultural sensor entities form a framework of the agricultural internet of things. The example creates an agricultural internet of things domain ontology based on semantic technology to describe resources, data, and spatiotemporal relationships in agriculture. The agricultural ontology knowledge model forms the basis of knowledge reasoning in the whole agriculture, so that concepts such as each sensor or intelligent equipment and the like share semantic knowledge. Agricultural ontology model as an example fig. 4 (containing part of the ontology model in agriculture):
(2) Semantic mapping:
and performing semantic mapping on the classification labels in the farmland model and the attributes in the body, wherein the semantic mapping is as shown in figure 3. The classification label "wet", "drought", "moderate" after machine learning is mapped with the attribute value of the soil humidity state in the ontology, the classification label "wet" is mapped as "wet" in the soil humidity state in the ontology, the classification label "drought" is mapped as "drought" in the soil humidity state in the ontology, and the classification label "moderate" is mapped as "moderate" in the soil humidity state in the ontology. The 'humidity', 'drought', 'moderate' after semantic mapping is not a label any more, but an abstraction of the soil humidity state on the semantic level has a certain semantic meaning, and based on the ontology, equipment in the farmland can understand the meanings of 'humidity', 'drought', 'moderate', and data sharing and intercommunication are achieved.
3) Ontology update and reasoning
And detecting farmland environment data in real time by using sensor equipment in the farmland, classifying and predicting the current environment data by using the farmland model trained in the previous step, and then performing semantic mapping. And after semantic mapping, updating the agricultural ontology by using the mapped semantic concepts, and finally reasoning based on the updated agricultural ontology to reason out agricultural events. For example, the environmental data at this time is predicted as a "drought" label, and then the "drought" label is mapped to a "drought" attribute value in the soil environment state in the ontology, and is updated to the ontology model, where the soil environment state has a "drought" attribute value. Based on the updated ontology, an "irrigation" event is derived by inference. Such as inference rules: (a rdf: type farmland) and (a hasSoilHumidityState b) and (b hasvalue "drought") → (a has _ treatment "irrigation");
as shown in fig. 8, the semantic mapping, ontology updating and reasoning process may include:
monitoring current farmland environment data;
classifying and predicting the detected farmland data to obtain a classification label;
semantic mapping is carried out on the classification result and concepts in the agricultural ontology model, so that semantic information can be obtained by the agricultural ontology knowledge model;
updating the agricultural ontology by using the mapped semantic concepts, which is equivalent to updating the directory state currently recorded by the knowledge model of the agricultural ontology;
and the agricultural ontology knowledge model decides the agricultural operation event.
In active learning, a sample selection method is a key influencing the accuracy of active learning, and a good sample selection method can select a sample with larger information quantity and representing the future development trend, so that the classification accuracy of an active learning model is continuously improved and has certain adaptivity. The present example provides a method for selecting a sample by comprehensively using clustering information of historical data and classification accuracy information of current data, and a specific implementation flowchart of the method is as shown in fig. 9:
step 1: first a sample selection Threshold is calculated.
Firstly, clustering historical sample data in passive learning according to classification labels, and calculating a sample center point c of each class i i ，
i is a classification category, X j Is the jth sample in the i class, i having n samples in total, c i Is the cluster center of the ith category. Then all samples in each class are calculated to the cluster center c i Of (d) is the mean value of the Euclidean distances i ，/>
Calculate all samples within each class to cluster center c i The maximum value D of the distance of (D),
let the initial value of the Threshold for sample selection be Threshold = d i +(D-d i ) X p, (sample selection threshold = cluster mean distance + (maximum distance of sample from cluster center-cluster mean distance) × classification accuracy of current sample), p the classification accuracy of the current new sample is set as 0 as the initial value, and then p Will change with the continuous selection of the sample, which will be given later p The calculation formula of (2). The clustering diagram is shown in FIG. 2. The new sample here corresponds to the aforementioned second sample data.
Step 2: selecting a sample according to a sample selection Threshold for manual marking;
when new sample data X is generated in the farmland, classifying by using the learned farmland model, wherein the classification label is i, and then calculating the distance d between the sample data X and the classification class clustering center thereof i 。
When the distance d i And when the Threshold value is less than or equal to the Threshold value corresponding to the classification, classifying by the farmland model, then performing semantic mapping and ontology updating on the labels and the ontology, and finally automatically reasoning out farmland events for operation.
When the distance d i When the value is larger than or equal to the Threshold value corresponding to the classified Threshold, feeding the sample data and the classification label back to agricultural experts for manual labeling, calculating the classification accuracy rate p of the farmland model,
and 3, step 3: updating a model according to the manual labeling condition, and updating Threshold;
if the manually labeled tags are consistent with the farmland model classified tags, the classified tags are directly subjected to semantic mapping and body updating, and finally, farmland events are automatically deduced for operation.
If the manually marked labels are inconsistent with the labels classified by the farmland model, semantic mapping and body updating are carried out by utilizing manual marking, and finally, farmland events are automatically deduced for operation. Meanwhile, the farmland model is iteratively updated by utilizing the artificially newly marked sample based on the stochastic gradient descent algorithm, and the clustering average distance d is updated by utilizing the new sample i The maximum distance D of the sample from the cluster center. According to the updated d i D and p updating a Threshold value of the sample selection, threshold = d i +(D-d i ) X.p. And then, circularly executing the steps to realize the self-adaptive automatic control of the agricultural Internet of things.
And finally, deleting samples which are long in time and far away from the clustering center after a period of time, retraining the farmland model, and then updating the model based on active learning.
In a word, aiming at the characteristics of diversity and variability of agricultural environment and crop habits, an active learning mechanism suitable for the characteristics of the agricultural Internet of things is designed based on the active learning idea. The active learning mechanism comprehensively utilizes the clustering information of historical data and the classification accuracy rate information of current data to select samples and update the model in an iterative manner, so that the model can change along with the changes of agricultural environment and crop habits, and is more suitable for the current agricultural environment.
The example combines active learning and semantic technologies to design a set of adaptive agricultural internet-of-things decision and automation control system. The system combines active learning and semantic technology, enables classification information not to be a label of '0' or '1' but to have certain semantic knowledge through semantic mapping, updates an agricultural ontology by using the classification knowledge learned through the active learning, and finally infers a crop planting scheme by means of abundant entities in the updated ontology and relations among the entities through reasoning.
The present embodiment further provides an electronic device, including:
a memory for information storage;
and the processor is connected with the memory and is used for realizing the data processing method provided by one or more of the technical schemes by executing the computer executable instructions stored on the memory, for example, the method shown in fig. 1, fig. 3, fig. 8 and fig. 9 can be executed.
A computer storage medium is provided in this embodiment, and the computer storage medium stores computer executable code; after the computer executable code is executed by the processor, the data processing method provided by one or more technical schemes can be realized; for example, the methods illustrated in fig. 1, 3, 8, and 9 may be performed. Alternatively, the computer storage medium may be a non-transitory storage medium.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above-described device embodiments are merely illustrative, for example, the division of the unit is only a logical functional division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another system, or some features may be omitted, or not implemented. In addition, the coupling, direct coupling or communication connection between the components shown or discussed may be through some interfaces, and the indirect coupling or communication connection between the devices or units may be electrical, mechanical or other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, all the functional units in the embodiments of the present invention may be integrated into one processing module, or each unit may be separately used as one unit, or two or more units may be integrated into one unit; the integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
Those of ordinary skill in the art will understand that: all or part of the steps of implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer-readable storage medium, and when executed, executes the steps including the method embodiments; and the aforementioned storage medium includes: various media capable of storing program codes, such as a removable Memory device, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A data processing method, comprising:
based on a classification model, obtaining first clustering information of first sample data, and determining a selection threshold according to the first clustering information;
acquiring second clustering information of first monitoring data of a target based on the classification model, and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold;
performing active learning of the classification model based on the second sample data, and optimizing model parameters of the classification model;
classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
and mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and obtaining an event decision for executing a preset operation on the target based on the updated body.
2. The method of claim 1,
the obtaining of the first clustering information of the first sample data based on the classification model includes:
based on the classification model, obtaining the clustering average distance of the first sample data, the maximum distance between the sample and a clustering center and the classification accuracy;
the determining a selection threshold according to the first clustering information includes:
and calculating the selection threshold according to the clustering average distance, the maximum distance between the sample and the clustering center and the classification accuracy.
3. The method of claim 2,
the obtaining of the second classification information of the first monitoring data based on the classification model includes:
clustering first monitoring data based on the classification model to obtain a clustering distance between the first monitoring data and a corresponding clustering center;
selecting second sample data from the first monitoring data according to the second classification information and the selection threshold, wherein the selecting of the second sample data comprises:
selecting, from the first monitoring data, a portion of the cluster distance greater than the selection threshold as the second sample data.
4. The method of claim 3,
the second classification information further includes: obtaining a classification label of the second sample data by using the classification model;
the method further comprises the following steps: acquiring a manual labeling label manually labeled by the second sample data;
and if the classification label of the second sample data is different from the artificial labeling label, actively learning the second sample data and the artificial labeling label by the classification model so as to optimize the model parameter of the classification model.
5. The method of claim 4, further comprising:
updating the classification accuracy of the classification model according to the comparison result of the classification label of the second sample data and the manual labeling label;
obtaining third classification information of the first sample data and the second sample data by using the updated classification model;
and updating the selection threshold according to the updated classification accuracy and the third classification information.
6. The method of claim 4, further comprising:
if the classification label of the second sample data is different from the manual labeling label, performing semantic mapping on the manual labeling label to obtain semantic information of the manual label; updating the ontology knowledge model based on the semantic information of the artificial tag, and performing event decision of executing a predetermined operation on the target by the ontology knowledge model based on the semantic information of the artificial tag;
or,
if the classification label of the second sample data is the same as the manual labeling label, performing semantic mapping on the manual labeling label or the classification label of the second sample data to obtain semantic information of the second sample data; updating the ontology knowledge model based on the semantic information of the second sample data, and performing event decision of executing a predetermined operation on the target by the ontology knowledge model based on the semantic information of the second sample data.
7. The method of claim 1, further comprising:
if the first monitoring data is not selected as the second sample data, performing semantic mapping on at least the classification label in the second classification information to obtain semantic information of the first monitoring data;
and updating the ontology knowledge model based on the semantic information of the first monitoring data, and performing event decision for executing a predetermined operation on the target by the ontology knowledge model based on the semantic information of the first monitoring data.
8. A data processing apparatus, characterized by comprising:
the threshold value determining module is used for acquiring first clustering information of the first sample data based on the classification model and determining a selection threshold value according to the first clustering information;
the selection module is used for acquiring second clustering information of the first monitoring data of the target based on the classification model and selecting second sample data from the first monitoring data according to the second clustering information and the selection threshold;
the active learning module is used for actively learning the classification model based on the second sample data and optimizing the model parameters of the classification model;
the classification module is used for classifying the second monitoring data of the target by using the optimized classification model to obtain a classification label of the second monitoring data;
and the mapping decision module is used for mapping the classification label of the second monitoring data into semantic information, updating the body by utilizing the semantic information, and obtaining an event decision for executing a preset operation on the target based on the updated body.
9. An electronic device, comprising:
a memory for information storage;
a processor coupled to the memory for implementing the method provided in any one of claims 1 to 7 by executing computer-executable instructions stored on the memory.
10. A computer storage medium having computer executable code stored thereon; the computer executable code, when executed by a processor, is capable of implementing the method as provided by any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811244716.0A CN111090707B (en) | 2018-10-24 | 2018-10-24 | Data processing method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811244716.0A CN111090707B (en) | 2018-10-24 | 2018-10-24 | Data processing method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111090707A CN111090707A (en) | 2020-05-01 |
| CN111090707B true CN111090707B (en) | 2023-03-31 |
Family
ID=70392442
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811244716.0A Active CN111090707B (en) | 2018-10-24 | 2018-10-24 | Data processing method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111090707B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115101058A (en) * | 2022-06-17 | 2022-09-23 | 科大讯飞股份有限公司 | Voice data processing method and device, storage medium and equipment |
| WO2024098174A1 (en) * | 2022-11-07 | 2024-05-16 | Qualcomm Incorporated | Model monitoring using input samples |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6874757B2 (en) * | 2016-02-24 | 2021-05-19 | 日本電気株式会社 | Learning equipment, learning methods and programs |
| CN108363716B (en) * | 2017-12-28 | 2020-04-24 | 广州索答信息科技有限公司 | Domain information classification model generation method, classification method, device and storage medium |
| CN108596266A (en) * | 2018-05-02 | 2018-09-28 | 深圳市易成自动驾驶技术有限公司 | Blending decision method, device based on semi-supervised learning and storage medium |
-
2018
- 2018-10-24 CN CN201811244716.0A patent/CN111090707B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN111090707A (en) | 2020-05-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20230263110A1 (en) | Turf management systems and methods | |
| US11256999B2 (en) | Methods and systems of forecasting the drying of an agricultural crop | |
| CN105005204A (en) | Intelligent engine system capable of automatically triggering intelligent home and intelligent life scenes and method | |
| CN111090707B (en) | Data processing method and device, electronic equipment and storage medium | |
| JPWO2020137085A1 (en) | Information processing equipment and information processing system | |
| US20190340485A1 (en) | Method and system for generating a responsive communication from a chatbot to network with plant monitoring systems | |
| Dighe et al. | Survey of crop recommendation systems | |
| Vaishnnave et al. | An empirical study of crop yield prediction using reinforcement learning | |
| Tripathy et al. | Smart farming based on deep learning approaches | |
| CN116090902A (en) | Remote farmland monitoring method and system | |
| CN112347264B (en) | Agricultural pest control method, device and computer readable storage medium | |
| Srichandan et al. | Data science and analytic technology in agriculture | |
| CN112347264A (en) | Agricultural pest control method and device and computer readable storage medium | |
| Chollet et al. | IoT-Enabled Agroecology: Advancing Sustainable Smart Farming Through Knowledge-Based Reasoning. | |
| CN117706954B (en) | Method and device for generating scene, storage medium and electronic device | |
| TWI780669B (en) | Method for managing machine learning model and computer readable medium therefor | |
| US20240028794A1 (en) | Event processing and prediction updating at a digital twin | |
| Nalla et al. | Green IoT and Machine Learning for Agricultural Applications | |
| WO2023042284A1 (en) | Crop growth assistance device, crop growth assistance method, and crop growth assistance program | |
| Kayum et al. | Appropriate Crop Recommended System for Cultivation using IoT and ML | |
| CN117557397B (en) | Method and system for controlling disinfection based on intelligent AI monitoring of warehouse pests | |
| CN117708680B (en) | Method and device for improving accuracy of classification model, storage medium and electronic device | |
| Balakrishnan et al. | Deep Reinforcement Learning with Fuzzy Inference System for Prediction of Crop-sowing Windows | |
| Paul et al. | Use of Artificial Intelligence in statistical research | |
| Zhai | Contributions to case-based reasoning enabled decision support system for smart agriculture |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |