Prerpints.org logo
Preprint
Article

Finding Negative Associations from Medical Data Streams Based on Frequent and Regular Patterns

Altmetrics

Downloads

97

Views

39

Comments

0

Submitted:

17 February 2024

Posted:

19 February 2024

You are already at the latest version

Alerts
Abstract
Medical data flows in streams as the data related to clinical tests and administered drugs by the doctors flows continuously. The doctors must be immediately alerted if negative associations are found among the drugs they prescribe. Data streams are to be processed in single scans as it is not possible to re-scan the data for any iterative processing. To detect negative drug connections, regular and frequent drug patterns must be processed. Negative correlations between disease-curing medications might create adverse responses that kill patients. This paper proposes an algorithm that finds the negative associations among regular and frequent patterns mined from medical data streams. The negative associations mined are the most effective and reduce the negative associations, which are more critical by 50%.
Keywords: 
Subject: Computer Science and Mathematics  -   Computer Science

1. Introduction

Association rule mining is a prominent data mining method for object or item set relationships [1,2,3]. Today, association rule mining requires massive data. The popular association rule mining method is Apriori [4].
Most of the medical data is generated continuously and moves in streams. Doctors carry diagnoses and prognoses. Continuously prescribes the medicines. The doctors need to be altered instantly when negative associations are found among the drugs prescribed by them.
Many Applications in modern days require capturing continuous data generated by man-held wearable devices. The data flowing through those streams must be processed, and any negative associations must be reported to the people concerned. Since data is detected and transported quickly, instream data is huge. Unlike databases, since instream data is never kept raw, it can only be scanned once.
Sensors and other streaming devices collect data now. The processed findings are briefly kept for analysis and decision-making, not the instream data. The instream data must be processed sequentially to uncover regular and frequent patterns with negative relationships.
Positive association rule mining is used for online log data, census data, biological data, fraud detection, and more. The number of negative associations decreases as the frequency of the Item set increases. Negative association rule mining can help improve healthcare crime data analysis decision support systems [5]. A basic technique mines too many patterns, creating too many association rules in which consumers may not be interested [6].
Di-Sets helps a vertical data presentation, which helps in comparing candidate patterns and reduces vertical table entry memory Mohammed J. Zaki [7].
Although FP-Tree [8] and RP-Tree are more stable and effective, many algorithms have been described. Several methods were developed to find sequential, irregular, unusual, etc. patterns. There’s little focus on uncovering rare negative patterns. Real-time settings are important for finding negative patterns, such as penguins being birds but not flying.
Patterns are item sets that occur in multiple transactions. Processing enormous numbers of patterns will take time. As the database grows, so do patterns. Frequent patterns matter more. Selecting patterns that fulfil the user-specified minimum threshold value finds frequent patterns. The user-selected support value is intriguing. The pattern frequency is how often it appears in the database. No timeframe is set here. Sporadic frequent patterns may not have a regular occurrence behavior.
Regular patterns can be inconsistent with frequent ones. Sometimes, both must be considered. Association between mined patterns is crucial. Positive connections between regular or regular-frequent, regular-frequent-maximal-closed patterns are commonly considered.
Some patterns are negatively correlated, meaning they conflict. When dealing with a medicine with a different chemical composition or weather forecast, one pattern contradicts another, leading to incorrect conclusions and actions. Negative associations are sometimes more essential than good ones and require clear examinations.
Based on the data to be processed, pattern discovery varies substantially. Streamed data must be handled as it flows and is not stored, although non-streamed data can be processed multiple times. Data flows are recognized, and actions are taken when recognized. Data streams are dynamic, with many variables and complicated objects. Such an environment benefits from the sliding window technique.
Mining common items and favorable relationships dominated the research. In addition to item set frequency, item regularity is significant. Negative associations show patterns of enemies, whereas positive relationships are desirable. Stream data flow is important, especially for medical data.
  • Problem Definition
Negative associations among chemicals of various drugs are dangerous as they lead to adverse effects on the patients. Medical data moves in streams continuously from the hospitals, which needs to be processed soon after the data is captured, and the negative associations are to be found and reported immediately. The item sets which are frequent and regular, which meet threshold criteria, must be found, and see if the item sets form any negative associations.
  • Objectives of the Research
  • Create a database and example set to investigate negative associations between regular, frequent, closed, and maximal items.
  • To create a mechanism for processing medical institution data streams.
  • Develop an algorithm to identify frequent, regular item sets and negative correlations among them.
  • Determine optimal frequency and regularity thresholds for accurate negative connections.

2. Related work

Many techniques for mining frequent data sets from static databases are published. Data flows in streams with the internet, especially from remote sensors to cloud storage. Dynamic mining follows data streams. To satisfy the user, respond quickly to constant requests. Time-sensitive sliding windowing was utilized by Chih-Hsiang Lin et al. [9] for frequent data stream mining. Their algorithm stores all frequent items in memory and a database containing expired data items. Based on storage space, the table can hold fewer expired data items.
One confidence threshold is typically used to identify positive and negative association rules (A ⇒ B, A ⇒ ¬B, ¬A ⇒ B, and ¬A ⇒ ¬B). In this method, the user must choose between positive and negative connections. X. Dong et al. [(CH4 ⇒ CH8).[10] compared confidence criteria for four positive and three negative associations. The linkages between the four confidences showed that four confidence intervals should focus on the four types of associations. Considering the four confidences helps create deceptive rules. They explained how the chi-squared test mines association rules. They suggested a chi-squared test-based PNARMC algorithm with four confidence thresholds.
Many data-gathering technologies have allowed every organization to store data on every activity. Large amounts of data make it hard to mine relevant information. Large data sets allow for more realistic and informative patterns. Sequential patterns give intriguing data insights. Existing sequential mining approaches focus on positive pattern behavior to anticipate the next event after a series. Fahad Anwar et al. [11] mine negative sequential patterns that contradict each other. Their technique finds patterned events/event sets.
Due to internet application migration, most data flows in streams. Streamed data is difficult to find fascinating. Support metric-based streamed data pattern mining is successful. However, pattern occurrence frequency is not a good measure for finding significant patterns. However, temporal regularity is best for online data stream mining for stock market applications. A pattern is regular if it appears within the user-defined period. None of the methods mine frequent patterns from online application stream data. Regular Pattern Stream tree (RPS tree) and an effective mining method for detecting interesting patterns in streamed data were developed by Syed Khairuzzaman Tanbeer et al. [12]. They recorded stream data in the RPS tree using a sliding window approach. The tree has been updated with the current data using an efficient approach.
Shirin Mirabedini et al. [13] reviewed all data stream pattern mining algorithms. Frequent item mining helps cluster and classify data. As network data volume has expanded, so needs to mine data streams for interesting patterns. Regular itemset mining is needed for static and streaming data. Frequent patterns might reveal business trends, scientific phenomena, etc. Pattern finding is the foundation for machine learning activities like association rule induction. Data has been scanned several times to uncover intriguing patterns. Only one scan of streamed data is possible; therefore, pattern-finding is performed in one scan. Since the stream length is unknown, such data has no closure. Before processing streaming data, an initial data set is collected and kept temporarily. VE: A review of online frequent pattern mining methods by Lee et al. [14]. They categorised approaches by pattern, data, and time window.
DSM-Miner by Juni Yang et al. [15] mines maximal common patterns effectively. Shirin Mirabedini et al. [13] examined all data stream pattern mining techniques. Frequent item mining clusters and classifies data. Mining network data streams for intriguing patterns has increased as data volume has grown. Static and streaming data require consistent itemset mining. Frequent patterns may reflect business, scientific, etc., trends. Machine learning operations like association rule induction start with pattern detection. Data was scanned multiple times to find unusual patterns. Only one streamed data scan is possible; therefore, pattern-finding is done there. No closure exists for such data because the stream length is uncertain. A temporary data set is collected before processing streaming data. VE Lee et al. present a structured review of online frequent pattern mining [14]. By pattern, data, and period, they classed approaches. Included the transaction sliding window mechanism that leverages each processing phase’s transaction count. Decaying distinguishes old and new transactions. SWM trees (Sliding window maximum frequent pattern trees) are suggested for retaining frequent patterns. SWM tree root is used as an enumerated tree root for searching.
Low interest or great confidence can make association rules stand out. The Daly et al. [16] method assessed exceptional mining rules. They explored negative and exceptional association rules. Negative association rules generate exceptions. Also, they created a metric for anomalous rule interest. Candidate rules analyze patterns and decisions using exceptional rules and metrics.
Most literature-based tactics cut desirable decision-making patterns utilizing fascinating criteria. However, choosing an interest measure is tricky and may need trial and error. No accurate method exists for determining interesting metrics. Thiruvady 2004[17] suggested using user inputs to determine rules and limitations. The GRD algorithm finds the most intriguing rules.
Statistical correlations determine data set connectivity. Maria-Luiza and Antonie [18] used a correlation between two item sets to identify negative association rules. Negative rules are retrieved if item sets correlate negatively, and confidence is strong.
Negative rules are retrieved if item sets are negatively correlated, and confidence is strong. Chris Cornelis [19] examined many algorithms that mine negative and positive association rules and identified various failures. These characteristics were used to classify and catalogue mining algorithms and identify gaps. A confidence framework-based modified Apriori mining method can find negative correlations with intriguing ones. They used upward closure to match validity definitions’ support-based negative connection interest. Dataset entries frequently have the intriguing “Support” parameter. Each level of data records has support values. Each level has multiple support values. The authors devised an Apriori-based upward closure method (PNAR) to find negative association rules. If ℸ X passes the minimal support condition, so do Y ⊆ I, X ∩ Y = ϕ , and ℸ(XY).
Interesting negative and positive association rules (PNAR) and mining multiple-level support approaches have been developed. Their method mines positive and negative association rules from fascinating frequent and infrequent item sets using varying support levels. Based on item set regularity, Pavan et al. [20] found positive and negative associations using vertical table mining.
Bagui and Dhar et al. [21] demonstrated how to mine positive and negative association rules from MAP REDUCE data. They utilised the Apriori approach to mine several item sets, which was efficient but required much computation time.
None of the negative association rule mining studies [22,23] used big data. Positive and negative association rules were found in uncommon item sets. The positive association rule mining identifies commonalities. This could reject many important or low-support items. Rare commodities or item sets can trigger negative association rules despite low support. Significant negative association rule mining requires more search space than positive rule mining for low-support objects. It would make sequential Apriori algorithm implementations easier and harder on big data. Few times has negative association rule mining been used.
Simarjeet Kauri et al. [24] reviewed AI-based illness diagnosis. However, no review has examined the medications’ effects on heart disease prediction. Jianxiong et al. [25] developed a machine learning-based drug response risk prediction algorithm. They anticipated disease-treatment drug risks. They have not considered patient risk when administering adversely related medications.
Lu Yuwen, Shuyu et al. [26] suggested “Prefix-Span” sequential data mining and “Proportional Report Ratio” disproportionality-based technique to detect major adverse drug responses based on casual relationships, drugs, and drug reactions. They tested single drug-to-drug reactions. Restricting adverse pharmaceutical responses has not been studied.
Yifeng Lu et al. [27] found predicted patterns in frequent item set mining. Negative correlations between suggested medicines are discovered. However, an uncommon item collection with negative drug connections is also important. The authors used bi-directional traversal to mine uncommon, closed item sets. However, negative relationships between infrequent or frequent item sets have not been studied.
Jingzhuo Zhang et al. [28] developed a system to obtain drug interactions for patients with various conditions. They compiled a drug–drug interaction database from medical sources. Distance monitoring was used to extract drug–drug interactions. A transformer-based bidirectional encoder representation was employed to extract drug interactions. There is no modelling to classify positive or negative interactions.
Generalised tensor decomposition was used to create drug–gene–disease connections by Yoonbee Kim et al. [29]. Chemical structure and ATC code drug feature networks predicted drug–gene–disease relationships. They learned drug, gene, and sickness using a multi-layer perceptron neural network. They emphasised positive connections and disregarded unfavourable ones, especially among drugs.
Aan Ahmed Toor etc al., [30] have presented a data stream mining method that includes enforcing Privacy before the same is disseminated. The data is mined from the data streams, and privacy is preserved. They primarily focussed on the drift in data Flowing in the streams due to changes in the data flow context.
U. H. W. A. Hewage et al. [31] have conducted a literature survey of the methods used for privacy-preserving data streaming mining methods. They focus primarily on the accuracy of privacy-preserving accuracy and data stream mining accuracy. They have not considered the issue of response time.
In his thesis, Aref [32] focused on developing a data stream association rule mining algorithm among co-occurring events, which means the kind of data stream varies occasionally. The proposed algorithm incrementally mines association rules over data streams that carry varied data.
Nannan Zhang et al. [33] have recommended a method to mine sequential patterns from data streams. They have used the sliding window method to arrive at the latest data by updating the old data with new data. A Prefix tree generates sequential patterns. They contended that Negative associations can be mined using the Prefix tree.
Most of the presentation in the thesis concentrated on mining drift privacy preservation of the data structure variations flowing in the data streams. The sliding window technique is used to store the current state of the data and ignore the continuous data flowing in the data stream. The main drawback is the inability to process data streams as they appear, find negative associations considering even the historical data and report the same to the destination to take control of false treatments.

3. Establishing data streams

A medical database is created out of registrations and prescriptions and fetching the details from pharmaceutical companies regarding the chemical compositions of the drugs. 100,000 records have been used to create the database.
Ninety thousand records were stored into a flat file, then 10,000 were placed into another flat file using streamed data. The same server holds both flat files. This implies that a 90,000-record static database and a 10,000-record streamed database are examined. Updates to the streamed database consider one window..
The data collected over the WEB is placed. Into a window (Buffer slot at a time), while the nearest is filled with the data, the rarest window is copied into a database. The size of Each window is designed to hold 1000 records. 10 such widow slots are used to hold the instream data.
Figure 1 shows the data streaming method used to process ongoing and continuous data flow from medical establishments.

4. Streamed database algorithms for negative association finding based on item set regularity and frequency

Algorithm-A This technique generates negative patterns using sliding windowing. Data is scanned as received, and the initial data is copied into 10 windows, each window of size 1000 records. 10 such windows are buffered, after which, while the latest window is received, the oldest window is written to a database.
A concurrent process accesses window database data. The Horizontal table, which lists item details and sales transactions, is updated. The pattern generation algorithm finds transaction data in instream data and writes it to a memory buffer. A buffer large enough to hold 1000 transactions is allocated. After round-robin buffer scanning, the horizontal table was updated based on transaction table data.
From a historical database, algorithm A creates an inverted table. Instream processing and Inverted Table updates are done in algorithm B. Process C creates negative linkages between medications and their development chemicals.
  • Algorithm- Process-A
  • Read the support value to set pattern frequency and user-defined regularity thresholds.
  • Read historical DBMS data into memory, as indicated in Table 1.
  • Convert Table 2 data to vertical format; see Table 2.
  • Algorithm Process-B
    Preprints 99109 i001
Table 3. Records from the Data stream into the window buffer
Table 3. Records from the Data stream into the window buffer
P.SL. No. Transaction ID Patient Number Disease Drug Chemicals Drug Chemicals
1 T1 P100 DE1 DR1 CH1 CH2 CH3 NA NA DR2 CH4 CH5 CH9 CH10
T3 P100 DE3 DR5 CH2 CH3 CH7 NA NA DR6 CH13 CH14 CH15 NA
T5 P223 DE5 DR9 CH1 CH3 CH5 CH16 CH19 NA NA NA NA NA
4 T7 P937 DE7 DR11 CH2 CH3 CH7 CH11 NA DR12 CH12 CH13 NA NA
5 T9 P119 DE9 DR15 CH1 CH3 CH5 NA NA DR16 CH8 CH9 NA NA
6 T11 P1235 DE11 DR19 CH5 CH8 CH11 CH15 NA DR20 NA NA NA NA
8 T15 P4573 DE15 DR27 CH1 CH3 CH5 NA NA DR28 CH9 CH11 NA NA
10 T19 P10987 DE19 DR35 CH1 CH3 CH5 NA NA DR36 CH6 CH9 CH10 NA
Table 4. Inverted Table for Instream Data
Table 4. Inverted Table for Instream Data
Chemical Code Transaction Ids
CH1 TT1 TT5 TT9 TT15 TT19
CH2 TT1 TT3 TT7
CH3 TT1 TT3 TT5 TT7 TT9 TT15 TT19
CH4 TT1
CH5 TT1 TT5 TT9 TT11 TT15 TT19
CH6 TT19
CH7 TT3 TT7
CH8 TT6
CH9 TT1 TT9 T15 T19
CH10 TT1 TT19
CH11 TT11 TT15
CH12 TT7
CH13 TT3 TT7
CH14 TT3
CH15 TT3 TT11
CH16 TT5
CH19 TT5
Table 5. Updated Medical Data (Historical Data + Instream data) (P - Patient, DE - Disease, DR = Drug, CH - Chemical in the Drug)
Table 5. Updated Medical Data (Historical Data + Instream data) (P - Patient, DE - Disease, DR = Drug, CH - Chemical in the Drug)
P.SL. No. Transaction ID Patient Number Disease Drug Chemicals Drug Chemicals
1 T1 P100 DE1 DR1 CH1 CH2 CH3 NA NA DR2 CH4 CH5 CH9 CH10
T2 P100 DE2 DR3 CH4 CH5 CH6 NA NA DR4 CH10 CH15 NA NA
T3 P100 DE3 DR5 CH2 CH3 CH7 NA NA DR6 CH13 CH14 CH15 NA
2 T4 P223 DE4 DR7 CH5 CH8 CH10 NA NA DR8 CH11 CH15 NA NA
T5 P223 DE5 DR9 CH1 CH3 CH5 CH16 CH19 NA NA NA NA NA
3 T6 P749 DE6 DR10 CH4 CH5 CH16 CH19 NA NA NA NA NA NA
4 T7 P937 DE7 DR11 CH2 CH3 CH7 CH11 NA DR12 CH12 CH13 NA NA
5 T8 P119 DE8 DR13 CH5 CH8 CH11 NA NA DR14 CH12 CH14 CH15 NA
T9 P119 DE9 DR15 CH1 CH3 CH5 NA NA DR16 CH8 CH9 NA NA
T10 P119 DE10 DR17 CH2 CH3 CH7 CH8 NA DR18 CH13 CH14 CH15 NA
6 T11 P1235 DE11 DR19 CH5 CH8 CH11 CH15 NA DR20 NA NA NA NA
7 T12 P11 DE12 DR21 CH4 CH5 CH6 NA NA DR22 CH10 CH15 NA NA
T13 P11 DE13 DR23 CH2 CH3 CH7 CH8 NA DR24 CH13 CH14 CH15 NA
T14 P11 DE14 DR25 CH5 CH8 CH11 CH15 NA DR26 NA NA NA NA
8 T15 P4573 DE15 DR27 CH1 CH3 CH5 NA NA DR28 CH9 CH11 NA NA
T16 P4573 DE16 DR29 CH4 CH5 CH6 NA NA DR30 CH14 CH15 NA NA
9 T17 P8765 DE17 DR31 CH2 CH3 CH6 CH7 NA DR32 CH12 CH13 NA NA
T18 P8765 DE18 DR33 CH5 CH8 CH11 CH12 NA DR34 CH14 CH15 NA NA
10 T19 P10987 DE19 DR35 CH1 CH3 CH5 NA NA DR36 CH6 CH9 CH10 NA
T20 P10987 DE20 DR37 CH4 CH5 CH6 NA NA DR38 CH12 CH14 CH15 NA
T21 P10987 DE21 DR39 CH2 CH3 CH4 NA NA DR40 CH7 CH13 NA NA
T22 P10987 DE22 DR41 CH5 CH8 CH11 NA NA DR42 CH12 CH15 NA NA
T23 P10987 DE23 DR43 CH1 CH3 CH5 NA NA DR44 CH9 CH14 NA NA
Table 6. Updated Inverted Table along with filled-up frequency and Regularity.
Table 6. Updated Inverted Table along with filled-up frequency and Regularity.
Chemical Code Transaction Ids Maximum
Regularity (4)		 Minimum
Frequency (3)
CH1 TT1 TT5 TT9 TT13 TT17 TT21 4 6
CH2 TT1 TT3 TT7 TT11 TT5 TT9 6 6
CH3 TT1 TT3 TT5 TT7 TT9 TT11 TT13 TT15 TT17 TT19 TT21 2 11
CH4 TT1 TT2 TT6 TT10 TT14 TT18 TT19 4 7
CH5 TT1 TT2 TT4 TT5 TT6 TT8 TT9 TT10 TT12 TT13 TT14 TT16 TT17 TT18 TT20 TT21 2 16
CH6 TT2 TT5 TT6 TT10 TT14 TT15 TT17 TT18 4 8
CH7 TT3 TT7 TT11 TT15 TT19 4 5
CH8 TT4 TT8 TT9 TT11 TT12 TT16 TT20 4 7
CH9 TT1 TT5 TT9 TT13 TT17 TT21 4 6
CH10 TT1 TT2 TT4 TT10 TT17 7 5
CH11 TT4 TT7 TT8 TT12 TT13 TT16 TT20 4 7
CH12 TT7 TT8 TT15 TT16 TT18 TT20 7 6
CH13 TT3 TT7 TT11 TT15 TT19 4 5
CH14 TT1 TT3 TT8 TT11 TT14 TT16 TT18 TT21 5 8
CH15 TT2 TT3 TT4 TT6 TT8 TT10 TT12 TT14 TT16 TT18 TT20 7 11
Table 7. Pruned Inverted Table for Maximum Regularity and Minimum Frequency
Table 7. Pruned Inverted Table for Maximum Regularity and Minimum Frequency
Chemical Code Transaction Ids Maximum
Regularity (4)		 Minimum
Frequency (3)
CH1 TT1 TT5 TT9 TT13 TT17 TT21 4 6
CH3 TT1 TT3 TT5 TT7 TT9 TT11 TT13 TT15 TT17 TT19 TT21 2 11
CH4 TT1 TT2 TT6 TT10 TT14 TT18 TT19 4 7
CH5 TT1 TT2 TT4 TT5 TT6 TT8 TT9 TT10 TT12 TT13 TT14 TT16 TT17 TT18 TT20 TT21 2 16
CH6 TT2 TT5 TT6 TT10 TT14 TT15 TT17 TT18 4 8
CH7 TT3 TT7 TT11 TT15 TT19 4 5
CH8 TT4 TT8 TT9 TT11 TT12 TT16 TT20 4 7
CH9 TT1 TT5 TT9 TT13 TT17 TT21 4 6
CH11 TT4 TT7 TT8 TT12 TT13 TT16 TT20 4 7
CH13 TT3 TT7 TT11 TT15 TT19 4 5
  • Algorithm Process-C
    Preprints 99109 i002
Table 8. Drugs mapped to Chemicals with Negative Association
Table 8. Drugs mapped to Chemicals with Negative Association
“Chemical” “Associated
Drugs”		 “Chemical” “Associated
Drugs”		 “Chemical” “Associated
Drugs”		 “Chemical” “Associated
Drugs”
CC4 DH3 CC8 DG7 CC11 DG11
CC6 DG3 CC8 DG7 CC11 DG11
CC4 DG3 CC6 DG3 CC8 DG7 CC11 DG11

5. Experimenting and Results

5.1. Creation of data set

A database contains patent registration, diagnostic, patient-diagnosis, chemical, drug-chemical, quantity, and prescription codes. Hospitals have contributed 100,000 patient registrations and medications to the database. Each diagnosis, drug administration, and drug chemical makeup are included in an Example set. Data items in repeating groups are encoded and replaced with codes. We sort the sample set, compute the frequency and regularity of each item set, update the database, and import 100,000 Flat file entries. These records were processed using this paper’s algorithm. No standard data set exists with the data items needed to discover negative connections.

5.2. Results

The procedure was applied on the data above, yielding the following results.
  • Step-1
    Read the Historical Data Contained in a Database and add transaction IDs for each transaction. Details of the data fetched from the Database is shown in Table 1.
  • Step-2
    Construct an Inverted Table in vertical format for the Transactions shown in Table 8 and Table 2, along with the frequency of occurrence and the Regularity. Regularity is based on database record placements, and frequency is based on occurrence count.
  • Step-3
    Read the Data from the Input data stream and load the same into a Temporary buffer. After filling the Temporary buffer, transfer the Records to window-1 of the window buffer after the last window is written to the database. The Records are read from the window Buffer. Table 3 shows the records from the data stream
  • Step 4
    Develop an Inverted Table from the buffered data. Table 4 shows the Inverted Table for records read from the window buffer. Here, the issue of frequency and regularity has not been used. The data is retrieved from online transactions.
  • Step 5
    Update the Historical Inverted Table with the data in the inverted table relating to the data stream. The frequencies and the regularity are updated simultaneously. The updated Transaction and Inverted are shown in Table 5 and Table 6.
  • Step -6
    Prune the records which do not meet the Maximum Frequency and Minimum regulatory.
    Records that do not satisfy the user’s criterion are pruned using the maximum regularity and minimum frequency of 4 and 3, respectively. Table 3 reveals that Chemical codes CH2, CH10, CH12, CH14, and CH15 were trimmed because they do not fulfil the Regularity and Frequency criterion. Table 7 lists leftover records. By this criteria, 5 chemicals were pruned. Yellow records are trimmed because they don’t meet thresholds.
  • Step-7
    Make adverse chemicals. Find Negative Associations with Records with No Common Transactions (Nill Common Items). Application of record intersection Produces unfavourable associations like
    (CC4 ⇒ CC8, CC11), (CC4 ⇒ CC8), (CC6 ⇒ CC8), (CC4 ⇒ CC11), (ACC6 ⇒ CC11), (CC8 ⇒ CC4, CC6), (CC6 ⇒ CC8, CC11), (CC11 ⇒ CC4, CC6), (CC4, CC6), ⇒ CC8, CC11)
  • Step 8 Find Negatively Associated Drugs
    As stated in Table 8, map back the chemicals connected with the medications to determine the negatively associated pharmaceuticals.
    The unfavourable associations show that DG3 should not be used with DG7 or DG11 because they conflict.

6. Discussion

Data Analysis has been conducted considering different sample sizes of 20,000, 50,000 and 80,000 Examples. The analysis is carried out considering thresholds in respect of frequencies and Regularities. While the sample sizes are used for the historical database, the Instream data size is fixed at 10,000 examples. Table 9 shows the number of negative associations considering the sample size of 20000 and data stream size of 10,000 Examples.
Data gathering yielded 100,000 cases evaluated for 30,000 sample sizes. 50,000, and 70,000). We set regularity and support thresholds and the number of unfavorable Associations observed. Table 9 shows frequent, regular, negative connections from 30,000 samples varying in regularity between 2.0 and 3.0 and frequency between 1.5 and 2.0.
Table 10 displays the number of negative Associations with frequency 1.75 and regularity 2–3 for 20,000 and 10,000 data stream instances. Negative associations respond to regularity modifications in Figure 1, with frequency fixed at 1.75. Figure 2 demonstrates the variation in negative correlations with regularity varying and Frequency set at 1.75.
Table 11 illustrates the number of negative connections with Frequency 1.5 and Regularity 2–3 in 30,000 cases. The line graph of negative associations with 1.5 frequency and variable regularity is shown in Figure 2. Figure 3 shows the variation in negative associations with regularity varying and frequency set at 1.50.
The larger example set was analysed further. Table 12 illustrates the percentage of negative correlations per 1000 Examples for sample sizes (20,000, 50,000, 80,000), including 10,000 DataStream instances. Frequency varies.
From the above Table, it can be inferred that Considering a sample size of 50,000 examples, fixing the regularity 1.65 and frequency 1.25 gives the least and most effective negative associations 20 out of 50,000 could be identified.
Further analysis examines the effects of fixing regularity, altering frequency, and different sample sizes. Table 13 lists unfavourable relationships with Regularity = 1.50 and Sample Size = 30,000. Figure 4 exhibits negative association variation.
Table lists negative relationships with a regularity 1.65 and a sample size of 50,000. Negative connections dropped 73%. Figure 5 shows variances.

7. Conclusions and Future Scope

Medical data moves in streams, and doctors must be warned soon after prescription if any negative associations exist among drugs prescribed by them, or else there could be a devastating effect on the patients. Many patients died during the coronavirus time as no system existed at that time that warned doctors not to use some drugs in combination. The use of humidifier drugs killed many due to many side reactions that affected the human organs.
Finding minimum and effective negative correlations is critical when medical data is rapidly shared online. Critical items are emphasised. Minifying regular and common item sets from historical databases and data streams yields the most effective negative associations. Since the complete data set is unavailable for study, finding the maximal and closed data sets is impossible. Regularity at 1.5 and frequency at 1.75 yields the last and most effective negative associations.
This research identifies all negative correlations with a collection of medications given to patients. Finding positive associations can be challenging due to many negative ones. Every medicine is chemically made. Chemical mixtures can be harmful. Consider negative correlations among chemicals and decode them into medications to determine the most important ones. This research is mainly limited by the chemical composition of medications doctors give to treat disorders.
Research medication administration limits and large trends. Rare and new medicinal item sets must be investigated. Further research may reveal intriguing elements beyond frequency and support that directly fit to uncover effective negative associations. The method may find negative links in remote or incremental medical databases and streaming data.

References

  1. Aggarwal, C.C.; Yu, P.S. Mining associations with the collective strength approach. IEEE Trans. Knowl. Data Eng. 2001, 13, 863–873. [CrossRef]
  2. Aggarwal, C.C.; Yu, P.S. A new framework for item-set generation. In Proceedings of the Seventeenth ACM SIGACTSIGMODSIGART Symposium on Principles of Database Systems, PODS’98, Seattle, WA, USA, 1–4 June 1998; pp. 18–24.
  3. Agrawal R, Imielinski T, Swami A. Mining association rules between sets of items in large databases. In: ACM SIGMOD conference. New York City: ACM Press; 1993. p. 207–216.
  4. Agrawal R, Srikant R. Fast algorithms for mining association rules. In: VLDB 1994 proceedings of the 20th international conference on large databases. 1994. p. 487–499.
  5. Mahmood S, Shahbaz M, Guergachi A. Negative and positive association rules mining from text using frequent and infrequent itemsets. Sci World J. 2014;2014:1–11. [CrossRef]
  6. Ashok SAVASERE, A., OMIECINSKI, E., AND NAVATHE, S. 1998. Mining for strong negative associations in a large database of customer transactions. In Proceedings of the Fourteenth International Conference on Data Engineering. IEEE Computer Society, Orlando, Florida, 494–502.
  7. Zaki, M.J. Fast Vertical Mining Using Diffsets. In Proceedings of the KDD03: The Ninth ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Washington, DC, USA, 24–27 August 2003.
  8. Chuan Yao Yang1 Yuqin Li1 Chenghong Zhang2 Yunfa Hu1 “A Novel Algorithm of Mining Maximal Frequent Pattern Based on Projection Sum Tree,” IEEE Transactions. 2007.
  9. Chih-Hsiang Lin, Ding-Ying Chiu, Yi-Hung Wu, Arbee L. P. Chen, “Mining Frequent Itemsets from Data Streams with a Time-Sensitive Sliding Window,” in Society for industrial and applied mathematics, 2006.
  10. X. Dong, F. Sun, X. Han, R. Hou, “Study of positive and negative association rules based on multiconfidence and chi-squared test,” in ADMA06, in LNCS, vol.4093, Springer-Verlag, Berlin–Heidelberg, 2006, pp.100–109.
  11. Fahad Anwar, Ilias Petrounias, Tim Morris, Vassilis Kodogiannis, “ Discovery of events with negative behavior against given sequential patterns” -1-4244- 5164-7.
  12. Syed Khairuzzaman Tanbeer, Chowdhury Farhan Ahmed, and Byeong-Soo Jeong,” Mining Regular Patterns in Data Streams,” DASFAA 2010, Part I, LNCS 5981, pp. 399–413, 2010, Springer-Verlag Berlin Heidelberg 2010.
  13. Shirin Mirabedini, Mahdi Ahmadi Panah, Maryam Darbanian, “Frequent Pattern Mining in Data Streams,” Journal of Engineering and Applied Sciences 12(4); 857-863, 2014.
  14. VE Lee, R. Jin, G. Agarwal, Frequent Pattern Mining in Data Streams, Frequent Pattern Mining, DOI 10.1007/978-3-319-07821-2 9, © Springer International Publishing Switzerland 2014.
  15. Juni Yang, Yanjun Wei, Fenfen Zhou, “An efficient algorithm for mining maximal frequent patterns over data streams,” 7th International Conference on Intelligent Human-Machine Systems and Cybernetics IEEE 2015.
  16. Daly, O.; Taniar, D. Exception Rules Mining Based On Negative Association Rules. Lect. Notes Comput. Sci. 2004, 3046, 543–552.
  17. Thiruvady, D.R.;Webb, G.I. Mining Negative Association Rules Using GRD. In Proceedings of the Pacific-Asia Conference on Knowledge Discovery and Data Mining, Sydney, Australia, 26–28 May 2004; pp. 161–165.
  18. Antonie, M.; Zaiane, O.R. Mining Positive and Negative Association Rules: An Approach for Confined Rules.. In Proceedings of the 8th European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD04), Pisa, Italy, 20–24 September 2004; pp. 27–38.
  19. Cornelis, C.; Yan, P.; Zhang, X.; Chen, G. Mining Positive and Negative Association Rules from Large Databases. In Proceedings of the 2006 IEEE Conference on Cybernetics and Intelligent Systems, Bangkok, Thailand, 7–9 June 2006.
  20. Chongqing, China, 29–31 May 2012. 23. Kumar, N.V.S.P.; Rao, K.R. Mining Positive and Negative Regular Item-Sets using Vertical Databases. Int. J. Simul. Syst. Sci. Technol. 2016, 17, 33.1–33.4.
  21. Bagui, S.; Dhar, P.C. Positive and negative association rule mining in Hadoop’s MapReduce environment. J. Big Data 2019, 6, 75. [CrossRef]
  22. Kishor, P.; Porika, S. An efficient approach for mining positive and negative association rules from large transactional databases. In Proceedings of the 2016 International Conference on Inventive Computation Technologies (ICICT), Coimbatore, India, 26–27 August 2016.
  23. Ramasubbareddy, B.; Govardhan, A.; Ramamohanreddy, A. Mining positive and negative association rules. In Proceedings of the 5th International Conference on Computer Science and Education, Hefei, China, 24–27 August 2018; pp. 1403–1406.
  24. Kaur, S.; Singla, J.; Nkenyereye, L.; Jha, S.; Prashar, D.; Joshi, G.P.; El-Sappagh, S.; Islam, M.S.; Islam, S.M. Medical Diagnostic Systems Using Artificial Intelligence (AI) Algorithms: Principles and Perspectives. IEEE Access 2020, 8, 228049–228069. [CrossRef]
  25. Wei, J.; Lu, Z.; Qiu, K.; Li, P.; Sun, A.H. Predicting Drug Risk Level from Adverse Drug Reactions Using SMOTE and Machine, Learning Approaches. IEEE Access 2020, 8, 185761–185775. [CrossRef]
  26. Lu, Y.; Chen, S.; Zhang, H. Detecting Potential Serious Adverse Drug Reactions using Sequential Pattern Mining Method. In Proceedings of the 2018 IEEE 9th International Conference on Software Engineering and Service Science (ICSESS), Beijing, China, 23–25 November 2018; pp. 56–59.
  27. Lu, Y.; Seidl, T. Towards Efficient, Closed Infrequent Item set Mining using Bi-directional Traversing. In Proceedings of the 2018, IEEE 5th International Conference on Data Science and Advanced Analytics (DSAA), Turin, Italy, 1–3 October 2018; pp. 140–149.
  28. Zhang, J.; Liu, W.; Wang, P. Drug-Drug Interaction Extraction from Chinese Biomedical Literature, using distant supervision. In Proceedings of the 2020 IEEE International Conference on Knowledge Graph (ICKG), Nanjing, China, 9–11 August 2020; pp. 593–598.
  29. Kim, Y.; Cho, Y. Predicting Drug–Gene–Disease Associations by Tensor Decomposition for Network-Based Computational Drug Repositioning. Biomedicines 2023, 11, 1998. [CrossRef]
  30. Aan Ahmed Toor, Muhammad Usman, Farah Younas, Alvis Cheuk M. Fong, Sajid Ali Khan 3 and Simon Fong, Mining Massive E-Health Data Streams for IoMT Enabled Healthcare Systems, sensors, 2020, 20, 2131.
  31. U. H. W. A. Hewage• R. Sinha • M. Asif Naeem, Privacy-preserving data (stream) mining techniques and their impact on data mining accuracy: a systematic literature review, Artificial Intelligence Review (2023) 56:10427–10464. [CrossRef]
  32. Aref Faisal Mourtada, Mining Association Rules Events over Data Streams, A Thesis submitted to Concordia Institute for Information Systems Engineering (CIISE), 2017.
  33. Nannan zhang, Xiaoqiang ren, and Xiangjun dong , An Effective Method for Mining Negative Sequential Patterns From Data Streams, IEEE Access, Vol 11, 2023, pp 31842 – 31854. [CrossRef]
Preprints 99109 g001
Figure 1. Capturing, buffering and creating a database representing the data streams.
Figure 1. Capturing, buffering and creating a database representing the data streams.
Preprints 99109 g001
Preprints 99109 g002
Figure 2. Negative association Item sets @ Frequency (1.75) and (30000 Examples) and varying Regularity.
Figure 2. Negative association Item sets @ Frequency (1.75) and (30000 Examples) and varying Regularity.
Preprints 99109 g002
Preprints 99109 g003
Figure 3. association Item sets @ Frequency (1.50) and (30000 Examples) varying Regularity
Figure 3. association Item sets @ Frequency (1.50) and (30000 Examples) varying Regularity
Preprints 99109 g003
Preprints 99109 g004
Figure 4. Negative association @ regularity (1.50) (30000 Examples) varying the Frequency
Figure 4. Negative association @ regularity (1.50) (30000 Examples) varying the Frequency
Preprints 99109 g004
Preprints 99109 g005
Figure 5. Negative association Item sets @ Regularity (1.65) (Examples 50,000) varying the Frequency
Figure 5. Negative association Item sets @ Regularity (1.65) (Examples 50,000) varying the Frequency
Preprints 99109 g005
Table 1. Sample Medical Data Extracted from Historical Database (P - Patient, DE - Disease, DR = Drug, CH - Chemical in the Drug)
Table 1. Sample Medical Data Extracted from Historical Database (P - Patient, DE - Disease, DR = Drug, CH - Chemical in the Drug)
P.SL. No. Transaction ID Patient Number Disease Drug Chemicals Drug Chemicals
1 T2 P100 DE2 DR3 CH4 CH5 CH6 NA NA DR4 CH10 CH15 NA NA
2 T4 P223 DE4 DR7 CH5 CH8 CH10 NA NA DR8 CH11 CH15 NA NA
3 T6 P749 DE6 DR10 CH4 CH5 CH16 CH19 NA NA NA NA NA NA
5 T8 P119 DE8 DR13 CH5 CH8 CH11 NA NA DR14 CH12 CH14 CH15 NA
T10 P119 DE10 DR17 CH2 CH3 CH7 CH8 NA DR18 CH13 CH14 CH15 NA
7 T12 P11 DE12 DR21 CH4 CH5 CH6 NA NA DR22 CH10 CH15 NA NA
T13 P11 DE13 DR23 CH2 CH3 CH7 CH8 NA DR24 CH13 CH14 CH15 NA
T14 P11 DE14 DR25 CH5 CH8 CH11 CH15 NA DR26 NA NA NA NA
T16 P4573 DE16 DR29 CH4 CH5 CH6 NA NA DR30 CH14 CH15 NA NA
9 T17 P8765 DE17 DR31 CH2 CH3 CH6 CH7 NA DR32 CH12 CH13 NA NA
T18 P8765 DE18 DR33 CH5 CH8 CH11 CH12 NA DR34 CH14 CH15 NA NA
T20 P10987 DE20 DR37 CH4 CH5 CH6 NA NA DR38 CH12 CH14 CH15 NA
T21 P10987 DE21 DR39 CH2 CH3 CH4 NA NA DR40 CH7 CH13 NA NA
T22 P10987 DE22 DR41 CH5 CH8 CH11 NA NA DR42 CH12 CH15 NA NA
T23 P10987 DE23 DR43 CH1 CH3 CH5 NA NA DR44 CH9 CH14 NA NA
Table 2. Inverted Table along with filled-up frequency and Regularity.
Table 2. Inverted Table along with filled-up frequency and Regularity.
Chemical
Code		 Transaction Ids Maximum
Regularity (4)		 Minimum
Frequency (3)
CH1 TT5 TT13 TT17 TT21 4 4
CH3 TT5 TT13 TT17 TT19 TT21 2 5
CH4 TT2 TT6 TT10 TT14 TT18 4 5
CH5 TT2 TT4 TT6 TT8 TT10 TT12 TT13 TT14 TT16 TT17 TT18 TT20 TT21 2 13
CH6 TT2 TT6 TT10 TT14 TT17 TT18 4 6
CH7 TT3 4 1
CH8 TT4 TT8 TT12 TT16 TT20 4 5
CH9 TT5 TT13 TT17 TT21 4 4
CH10 TT2 TT4 TT10 TT17 7 4
CH11 TT4 TT8 TT12 TT13 TT16 TT20 4 6
CH12 TT8 TT16 TT18 TT20 7 4
CH13 TT3 4 1
CH14 TT8 TT14 TT16 TT18 TT21 5 5
CH15 TT2 TT4 TT6 TT8 TT10 TT12 TT14 TT16 TT18 TT20 7 10
Table 9. Negative Frequent Regular Associations (20,000 Historical and 10000 Data Stream Examples) Vs Varying Regularity + Frequency
Table 9. Negative Frequent Regular Associations (20,000 Historical and 10000 Data Stream Examples) Vs Varying Regularity + Frequency
Total Transactions “%Max Regularity” “%Support Count” “Number of
Negative Frequent
Regular assertions”
30,000 3 2 3
3 1.75 21
3 1.625 77
3 1.5 230
30,000 2.5 1.75 20
2.5 1.625 77
2.5 1.5 176
2.5 1.125 490
30,000 2 1.75 18
2 1.625 59
2 1.5 90
2 1.1.25 334
Table 10. Negative Associations vs. Frequency @ 1.75 and varying Regularity (20,000 historical + 10,000 data stream examples).
Table 10. Negative Associations vs. Frequency @ 1.75 and varying Regularity (20,000 historical + 10,000 data stream examples).
“Regularity” “Negative Frequent
Regular Item Sets”		 % of Negative Associations * 1000
3.00 3 3 / 30,000 = 0.10
2.50 20 20/ 30,000 = 0.66
2.00 18 18/30,000=0.60
The average percentage of Negative association per 1000 Examples 0.45
Table 11. Negative Associations @ Frequency at 1, (30000 Examples) Varying Regularity
Table 11. Negative Associations @ Frequency at 1, (30000 Examples) Varying Regularity
“Regularity” “Negative Frequent
Regular Item Sets”		 % of Negative Associations * 1000
3.00 230 230/30000 =7.6
2.50 176 176/30000 = 5.8
2.00 90 90/30000= 3.0
Average Negative Associations per 1000 Examples 5.4
Table 12. % reduction in negative relationships with larger samples and appropriate frequency and support
Table 12. % reduction in negative relationships with larger samples and appropriate frequency and support
“Total
Transactions”		 “%Max
Regularity”		 “%Support
Count”		 “Number of
Negative
Frequent
Regular”		 % of Negative
Associations * 1000		 Average % of
Negative
Associations
per 1000 samples
30,000 1.50 1.750 18 18/30=0.60 1.85
1.50 1.625 59 59/30=1.96
1.50 1.500 90 90/30=3.0
50,000 1.65 1.65 7 7/50=0.14 0.60
1.65 1.25 20 20/50=0.40
1.65 1.00 75 75/50=1.25
80,000 1.35 1.65 18 18/80=0.22 0.70
1.35 1.35 59 59/80=0.73
1.35 1.00 90 90/80=1.13
Table 13. Negative associations @ Regularity1.50 (30000 Examples) Varying frequency
Table 13. Negative associations @ Regularity1.50 (30000 Examples) Varying frequency
“Frequency” “Negative Frequent
Regular Item Sets”		 “% of Negative
Associations * 1000”
1.750 18 18/30=0.60
1.625 59 59/30=1.97
1.500 90 90/30=3.00
Average 1.85
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.
Prerpints.org logo

Preprints.org is a free preprint server supported by MDPI in Basel, Switzerland.

facebook logotwitter logolinkedin logo
MDPI Initiatives
Important Links
Subscribe

Disclaimer

Terms of Use

Privacy Policy

© 2024 MDPI (Basel, Switzerland) unless otherwise stated